SlideShare a Scribd company logo
1 of 79
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
2022 AI/ML and Data Edition
Chris Fregly
Principal Specialist Solution
Architect @ AWS AI/ML
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
A P P S
D E V I C E S
P E O P L E
A P P / L O G S
T H I R D - P A R T Y D A T A
I O T / D E V I C E S
Data sources
F O R
A P P L I C A T I O N S
Amazon
Aurora
Amazon Kinesis
& Amazon MSK
F O R A N A L Y T I C S A N D
M A C H I N E L E A R N I N G
Data Lake
Amazon S3
Amazon Redshift
Data Warehouse
Amazon
Redshift
Amazon
EMR
B U S I N E S S
I N T E L L I G E N C E
Amazon
QuickSight
M A C H I N E
L E A R N I N G
Amazon
SageMaker
A N A L Y T I C S
Amazon
DynamoDB
AWS Glue
| AWS Lake Formation, Amazon DataZone
Building an end-to-end ML and data strategy
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
ML infrastructure and hardware
ML and data governance
Discover, analyze, and prepare data
Build and train ML models
Deploy ML models for inference
Low-code / no-code ML
AI services
ML education
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
ML infrastructure and hardware
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Journey of silicon innovation at AWS
AWS Inferentia
and AWS Trainium
Machine learning acceleration
AWS Graviton
Powerful and efficient,
modern applications
AWS Nitro System
Hypervisor, Nitro Cards, network,
storage, SSD, and security
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker supports new instance types
SageMaker Model Training support for ml.trn1 instances
• Powered by AWS Trainium chips
• ml.trn1.2xlarge, for experimenting with a single accelerator
and training small models cost effectively
• ml.trn1.32xlarge for training large-scale models
SageMaker Inference adds eight new Graviton-based instances
• Powered by Graviton3 and Graviton2
• For Real-time and asynchronous inference model deployment options
• Graviton3: ml.c7g
• Graviton2: ml.m6g, ml.m6gd, ml.c6g,
ml.c6gd, ml.c6gn, ml.r6g, and ml.r6gd
Trn1
C7g M6g C6g R6g
p
r
e
:
I
n
v
e
n
t
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
New instance types for Amazon SageMaker
P4de instance
documentation
P4de
• Provide the highest performance for ML training and HPC
applications
• Powered by 8 NVIDIA A100 GPUs with 80 GB high-performance
HBM2e GPU memory, 2X higher than the GPUs in current P4d
instances
• Up to 640GB of GPU memory, providing up to 60 percent
better ML training performance along with 20 percent lower
cost to train when compared to P4d instances
SageMaker Model Training support for ml.p4de.24xlarge instances
Dec 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
ML and data governance
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Why ML and data governance?
9
Machine learning (ML) governance
Onboard Develop Monitor
User setup
ML activities Deployment
Build
Train
Prepare
Tune
Inferences
Customers
Business
applications
Platform admin Data engineer Data scientist ML engineer ML risk officer Model approver
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker – New ML Governance Tools
Simplify access control and enhance transparency
Amazon SageMaker
Role Manager
Define custom user
permissions in minutes
Amazon SageMaker
Model Cards
Centralize model information
and documentation
Amazon SageMaker
Model Dashboard
Monitor model performance
in one place
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Role Manager
Define custom user permissions in minutes
• Simplify permissions for
ML activities
• Use guided workflows for
role creation
• Accelerate user onboarding
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Model Cards
Easily document, retrieve, and share the necessary model information
• Streamline model documentation
• Capture model information, such
as input datasets, training
environments, training results,
model purpose, performance goals
• Attach and visualize evaluation
results, such as bias and quality
metrics
• Share model cards with business
stakeholders, internal teams, or
your customers
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Model Dashboard
Unified view across all your models to audit performance
• Track model behavior
• Integrates with SageMaker Model
Monitor and SageMaker Clarify
• Monitor model behavior for data
quality, model quality, bias drift,
and feature attribution drift
• Automate alerts
• Troubleshoot model deviations
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Coming Soon
Share data securely with Amazon DataZone
Unlock data across organizational boundaries with build-in governance
Amazon
DataZone
Data
producers
Data
consumers
Fine-grained controls to manage and
govern access to data
Discover and share data at scale
across organizational boundaries
Makes it easy for data scientists and
other business users to discover, use,
and collaborate around that data
Manage organization-wide
governance in one place
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
SageMaker Studio passes user permissions to EMR
Multiple SageMaker Studio users can connect to the same EMR cluster with isolated access
• All EMR jobs created from SageMaker
Studio will inherit data and resources
permissions for the given user.
• Multiple SageMaker Studio users can use the
same EMR cluster with separate access to data
• When accessing data lakes managed by Lake
Formation, table-level and column-level access
policies are enforced
p
r
e
:
I
n
v
e
n
t
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Dynamic data masking with Amazon Redshift
Protect sensitive data with role-based permissions
ID
Geo-
location Name Phone number
123 WA Ana 123-456-3568
124 NY Alice 123-457-****
125 WA Bruce 123-457-3569
126 CA Chris 123-457-****
130 CA Sharon 123-457-****
Condition
column
Mask column
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Granular access with Redshift and Lake Formation
Centrally manage data sharing in Amazon Redshift with AWS Lake Formation
Amazon
Redshift
Amazon
Redshift
Amazon
Redshift
AWS
Lake Formation
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Amazon Security Lake
Automatically centralize security data into a purpose-built data lake in a few steps
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Discover, analyze, and prepare data
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Integration with many popular data sources
Supports both AWS and 3rd-party data source integrations from Amazon SageMaker
• Amazon S3, Athena, Redshift, EMR
• Salesforce, ServiceNow, Marketo
• Mailchimp, SendGrid, Zendesk, Jira, Datadog
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
AWS Data Exchange for AWS Lake Formation
Secure data mesh architecture with third-party data using AWS Data Exchange
AWS Lake Formation
Providers use LF-tags
to indicate which
data subscribers
should have access to
Third-party data
provider
Provider stores data in
AWS Lake Formation
AWS Data Exchange
Grants subscriber read-
only access to the data
tagged with the key-value
pairs specified by the
provider when a
subscription starts and
automatically denies
access when it ends
AWS Marketplace
Automated payments
and billing
Subscriber
Subscribers see
the data once
their account
is verified
Query provider data
Query, transform, or
share access with the
appropriate user
groups without any
upfront ETL
AWS account
+
AWS Data Exchange for AWS Lake Formation
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Amazon CodeWhisperer code generator
Enterprise administrative controls, simple sign-up, and support for new languages
• Generates code recommendations based on
the comments – and prior code - in your IDE
• Available in popular IDEs such as Visual
Studio Code, JetBrains, AWS Cloud9, AWS
Lambda console
• Supports Python, Java, JavaScript, C#,
TypeScript
• Enable CodeWhisperer for your organization
with single sign-in (SSO) authentication
• Sign-up with AWS Builder ID
• Generates open source attribution
documentation for you
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
AWS Glue Data Quality continuous monitoring
Deliver high quality data across your data lakes and data pipelines
• Automatic data quality rule
recommendations based on your data
• Keep data quality high with ongoing data
analysis and quality checking
• Data quality for datasets in your data lake
and data pipelines
• Cost-effective to scale with pay-as-you-go
billing, with no lock-in
AWS Glue Data Quality
Amazon SageMaker
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Glue for Ray
Scaling your data integration workloads using Python
• AWS Glue for Ray is a new engine option on
AWS Glue.
• Data engineers and ML practitioners can use
AWS Glue for Ray to process large datasets
with Python and popular Python libraries.
• AWS Glue jobs are fire-and-forget systems
where you can submit your Ray code to the
AWS Glue jobs API
• AWS Glue for Ray facilitates the distributed
processing of your Python code over multi-
node clusters.
AWS Glue
Amazon SageMaker
Preview
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
• Access interactive Spark clusters that start in
under a second and run faster with our
optimized runtime for Apache Spark
• Harness Apache Spark for complex, powerful
analytics using the expressive power of
Python along with its wide ecosystem
• Build Apache Spark applications without
managing resources or configuring software
using Amazon Athena
Amazon Athena
Amazon Athena for Apache Spark
Run interactive analytics on Apache Spark in under a second
Amazon SageMaker
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Redshift integration for Apache Spark
Build Interactive Spark Applications with Amazon SageMaker, Glue, and EMR
Redshift
Connector for
Apache Spark
Amazon Redshift
Amazon Glue
Amazon EMR
• Apache Spark applications accessing Amazon
Redshift data from AWS analytics services such as
Amazon EMR, AWS Glue, and Amazon SageMaker
• Build Apache Spark applications that read from
and write to your Amazon Redshift data
warehouse, without compromising performance or
transactional consistency.
• No manual setup and maintenance of uncertified
versions of Spark-Redshift open-source connectors
• Improved performance with only relevant data
moved from Redshift to consuming applications
Amazon SageMaker
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Streaming data support for Amazon Redshift
Directly ingest streaming data into your data warehouse for real-time processing
• Directly ingest streaming data into Amazon
Redshift from Kinesis Data Streams and Managed
Streaming for Apache Kafka without staging in S3
• Perform rich analytics using familiar SQL on
streaming data
• Easily create and manage extract-load-transform
(ELT) pipelines with streaming data
• Process large volumes of streaming data from
multiple sources to derive insights in seconds
Amazon
SageMaker
Amazon Kinesis
Data Streams
Amazon Managed
Streaming for
Apache Kafka
Redshift
Kinesis or Kafka
producer
KDS Stream
MSK Topic
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Multi-AZ support for Amazon Redshift
Highly resistant data warehouse with auto-failover and no data loss
• Workload processing across availability zones
(AZs)
• Easy management through a single endpoint
• Auto-failover with zero data loss and no
manual intervention
Amazon Redshift
managed storage
AZ 1 AZ 2
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Amazon Redshift Auto-Copy from Amazon S3
Simplified and automated file ingestion from Amazon S3 into Redshift
• Simple, low code data ingestion
• Avoid re-ingestion and manual tracking of
loaded files
• Easily convert your existing COPY statements
into automatic ingestion jobs
• Automatic ingestion of new data from
Amazon S3 based on user defined
configurations
Amazon S3
Redshift
Copy Job
Redshift
Table
Continuously
monitoring S3
folder
New file(s)
detected Ingestion
automatically
starts
Amazon
SageMaker
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Zero-ETL integration from Redshift to Aurora
Access multiple Amazon Aurora databases with Amazon Redshift
• Drive holistic insights across applications
or partitions
• Analyze data from multiple Aurora
databases in the same Redshift cluster
• Leverage Redshift features such as
materialized views, data sharing and
federated access to data lakes
Amazon Redshift Amazon Aurora
Amazon
SageMaker
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Serverless EMR and Redshift are now GA
Pay only for the resources you use with these serverless options
Amazon EMR Serverless
• Run Spark and Hive applications without having to configure, optimize, or manage clusters
• Fine-grained auto-scaling of compute and memory resources
• Uses the performance-optimized EMR runtime
Amazon Redshift Serverless
• Run analytics queries without having to configure, tune, and manage data warehouse clusters
• Intelligently auto-scales data warehouse capacity to match your workload demand in seconds
• Supports Redshift Query Editor v2 or any business intelligence (BI) tool of your choice
Sum
m
er 2022
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Data Wrangler new features
Built-in data preparation in SageMaker Studio Notebooks
• Automatically generates key
visualizations on top of Pandas
data frames
• Understand data distribution and
identify data quality issues
• Generate ML-specific insights for
ML target column
• Receive recommendations for
data transformations and code
import pandas as pd
import sagemaker_datawrangler
df = pd.read_csv("data.csv")
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Data Wrangler new features
Deploy data preparation flows for real-time and batch inference
Data Wrangler
Flow
Data
Scientist
ML
Engineer
Amazon SageMaker Data Wrangler
Data
Preparation
Job
Model
Training
Inference
Pipeline
Run data
preparation
for model
training
Reuse data
transformation flow
for real-time & batch
inference
Define data
preparation
for training
Deploy
inference
• Deploy data preparation
flows from SageMaker Data
Wrangler for real-time and
batch inference
• Reuse the data
transformation flow
• Speed up your production
deployment
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker Data Wrangler + EMR/Athena
Now supports Amazon EMR Presto and EMR Spark as big-data query engines
• Connect to existing Amazon EMR
Presto and Athena clusters using a
visual experience in SageMaker Data
Wrangler
• Prepare data for ML in minutes using
Data Wrangler’s visual interface
• Analyze data, clean data, and create
features for ML using 300+ built-in
transformations backed by Spark
without the need to author Spark code
Amazon EMR
Presto
Amazon SageMaker
Data Wrangler
D
e
c
2
0
2
2
Amazon Athena
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Apache Iceberg: Amazon SageMaker Feature Store
Now supports Apache Iceberg table format
• Create feature groups in the offline
store in Apache Iceberg table format
• Apache Iceberg is an open table
format for very large analytic
datasets
• Apache Iceberg compacts small data
files into fewer large files in the
partition, resulting in significantly
faster queries.
Amazon SageMaker
Feature Store
Apache Iceberg
D
e
c
2
0
2
2
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Omics
Store, query, analyze, and generate insights from genomic and other omics data
• Built-in access control and logging
D N A R N A P R O T E I N S
Genomics Transcriptomics Proteomics
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Expanded API capabilities for Amazon QuickSight
Programmatic access to the underlying structure of QuickSight dashboards
• Access underlying data models of Amazon
QuickSight dashboards, reports, analyses and
templates via the AWS Software
Development Kit (SDK).
• Translate legacy BI assets to cloud-native
dashboards quickly
• Reduce migration time from months to
weeks
• Integrate into DevOps processes such as code
reviews, audits, and audit every change
before deployment
Amazon QuickSight
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon QuickSight Paginated Reports
Day-to-day detailed operational data in custom formats
• Create, schedule, and share highly formatted
multipage reports
• Build all insights, independent of preferred
consumption model, on single source of truth
governed datasets
• Single authoring experience for dashboards
and reports
• Unified consumption experience allows users
to choose dashboards and reports
• Pay for what you use.
Amazon QuickSight
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Build and train ML models
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Amazon CodeWhisperer code generator
Enterprise administrative controls, simple sign-up, and support for new languages
• Generates code recommendations based on
the comments – and prior code - in your IDE
• Available in popular IDEs such as Visual
Studio Code, JetBrains, AWS Cloud9, AWS
Lambda console
• Supports Python, Java, JavaScript, C#,
TypeScript
• Enable CodeWhisperer for your organization
with single sign-in (SSO) authentication
• Sign-up with AWS Builder ID
• Generates open source attribution
documentation for you
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Improved Amazon SageMaker Studio usability
Redesigned user interface (UI) and user experience (UX) based on customer feedback
• Redesigned navigation menu
following ML workflow
• Dynamic landing pages with links
to videos, tutorials, blogs, etc.
• New SageMaker Home page with
one-click access to common tasks
• Redesigned launcher with quick
links to create notebook, open
console, etc.
Supported for SageMaker Studio domains running on JupyterLab 3+
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Notebook Jobs for Amazon SageMaker Studio
Automatic conversion of notebook code to production-ready jobs
• Run your notebooks as is or in a
parameterized fashion with just a few
simple clicks from the SageMaker
Studio
• Run notebooks on a schedule or
immediately
• No need for the end-user to modify
their existing notebook code
• View the populated notebook cells
after the job is complete, including any
visualizations
• Also available in SageMaker Studio Lab
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Improved Amazon SageMaker Experiments
Next-generation experiment tracking, model comparison
• New UI for easier experiment tracking
and model comparison
• Organize, track, compare and evaluate
machine learning (ML) experiments
and model versions from any IDE,
including local Jupyter notebooks
• Programmatic tracking of experiments
and logging of metrics/parameters
Dec 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Collaboration with Amazon SageMaker Studio
Team-based sharing and real-time collaboration with shared spaces
• SageMaker Studio now supports
team-based sharing and real-time
collaboration
• Create shared spaces to access, read,
edit, and share the same notebook in
real time
• Shared EFS directory in a space
• Filtered SageMaker resources
Experiments
Model Registry
Pipelines, etc.
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Collaboration with Amazon SageMaker Studio
Team-based sharing and real-time collaboration on the same notebooks
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker + Glue Interactive Sessions
Scaling your data integration workloads using interactive Apache Spark and Ray
• From SageMaker Studio, access AWS Glue
Interactive Sessions to access large, serverless,
distributed clusters for Apache Spark and Ray
• Data engineers and ML practitioners can use
Apache Spark or Ray to process large datasets
with Python and popular Python libraries.
• Train distributed ML models on large datasets
using algorithms from Apache Spark or Ray
AWS Glue Interactive
Sessions
Amazon SageMaker
Preview
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Heterogeneous clusters for Training Jobs
Reduce training cost by using a mix of instance types specific to your workload
• Use both CPU-optimized and GPU-
optimized instance types in a
single distributed training job
• Perform data transformations on
CPU instance types
• Perform ML operations on GPU
instance types
O
ct 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Warm pools for SageMaker Training Jobs
Reduce startup time for SageMaker Training and Tuning Jobs with warm pools
• Enables faster iterations for ML
development lifecycle
• Up to 8x improvement in startup
times for SageMaker Training,
Tuning, and Autopilot jobs
• Balance cost and convenience with
configurable keep-alive times
Sept 2022
from sagemaker.pytorch import PyTorch
estimator = PyTorch(
entry_point='train.py’,
...
keep_alive_period_in_seconds=1800)
estimator.fit('s3://my_bucket/my_data/')
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
SageMaker supports more domains and users
Now supports multiple domains within the same AWS Account
• Create multiple SageMaker
domains within the same AWS
account
• Scope access and isolate resources
to different team or business units
in your organization
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker tagging and cost-allocation
Automated tagging for better cost allocation and monitoring
• SageMaker automatically tags
resources at domain, user and space
level supporting detailed cost
allocation
• All SageMaker resources with an Amazon
Resource Name (ARN) including:
• Training jobs
• Processing jobs
• Kernel Gateways
• Endpoints
• etc.
• Supports more detailed cost allocation for
administrators
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
• Bring your own or acquire geospatial
data with just a few clicks
• Easily prepare geospatial data with
built-in operations and
transformations
• Speed model building with pre-
trained deep neural network (DNN)
models and geospatial operators
• Analyze and explore predictions with
built-in visualization tools
Amazon SageMaker – Geospatial ML
Build, train, and deploy ML models using geospatial data
Aerial and
satellite imagery
Mapping
data
Road mask
(color as speed)
24 Hour Fitness
7-Eleven
76 Gas Stations
84 Lumber
85C Bakery Cafe
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Preview
Amazon SageMaker – Geospatial ML
Visualize predictions
Select or train a model
Access geospatial data sources
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Updates to Amazon SageMaker Pipelines
Supports local mode and cross-account artifact sharing
Fully supports “local mode” for iterative development and testing
• Fast feedback loops are critical to speed up pipeline development and testing
• Develop and test pipelines locally to reduce cost and increase developer efficiency
• Use on your laptop with any IDE such as VSCode, PyCharm, vi, or emacs
Securely share pipeline artifacts across multiple accounts
• Multi-account strategy helps achieve data, project, and team isolation
• Share pipeline artifacts between lines of business within your organization
• Supports different authorization levels including read-only and execute permissions
Fall 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Deploy ML models for inference
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Shadow testing for Amazon SageMaker Endpoints
Now supports shadow testing to compare models in real-time
• Validate performance of your models by
comparing them to production models
• SageMaker takes care of mirroring
requests
• Start small and dial up to control costs
• Catch potential configuration errors and
performance issues before they impact
end users
• Monitor progress of the shadow test and
performance metrics such as latency and
error rate through a live dashboard
Amazon SageMaker Endpoint
Production Variant
Shadow Variant
Model A
Model B
Request
Response
Request
Request
Response
Application
Response
Amazon S3
Accessible through AWS console, CLI, APIs
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
• Set up a test for a
predefined duration
• Monitor operational
performance through a live
dashboard
• Deploy models with
confidence or rollback
Shadow testing for Amazon SageMaker Endpoints
Tools to manage shadow tests
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
More support for large models and GPUs
Now supports deploying large models that require large volume size
• Deploy large models (up to 500GB) for inference on SageMaker’s Real-time and
Asynchronous Inference options by configuring the maximum EBS volume size and
timeout quotas
• Container health check and download timeout quotas have been made configurable
up to 60 minutes, so you have more time to download and load your model and
associated resources
Multi Model Endpoint (MME) now supports GPU-based instances
• Use MME to deploy thousands of ML models on GPU-based instances
• MME dynamically loads and unloads models from GPU memory based on incoming
traffic to the endpoint
• Save cost with MME as the GPU instances are shared by thousands of models
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Multi-model Endpoint (MME) support for GPUs
Use MME to deploy thousands of ML models on GPU-based instances
Run hundreds of models
behind a single endpoint to
maximize GPU utilization
Powered by NVIDIA
Triton Server to
support run models
trained on your choice
of popular frameworks
Wide selection of
up to 15 GPU
instance types
Automatically
applies endpoint
autoscaling policy to
all models
P3
G4 G5
🤗 Hugging Face
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Monitor Amazon SageMaker Batch Transform Jobs
p
r
e
:
I
n
v
e
n
t
Now supported for Batch Transform Jobs in Amazon SageMaker
• Monitor the quality of ML predictions from Batch Transform jobs in SageMaker
using Amazon SageMaker Model Monitor
• SageMaker Batch Transform enables you to run predictions on datasets stored in
Amazon S3
• Collect Batch Transform data in production, analyze it, and compare it against
your training or validation data to detect deviations
• Use SageMaker Model Monitor’s built-in rules to detect drift for structured data
sets, add data transformations before you run the built-in rules, or write your own
custom rules
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Low-code / no-code ML
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker low-code / no-code
Business
Requirements
Data Preparation &
Feature Engineering
Model Development,
Training, and Tuning
Model Deployment
Inference & Monitoring
Autopilot
AutoML capability that automatically prepares your data, as well as builds, trains, and
tunes the best machine learning models for your tabular and time-series datasets
JumpStart
Pre-built solutions and a model zoo of pre-trained and easily tunable state-of-the-art
models for Computer Vision, and Natural Language Processing
A dedicated workspace for data engineers, data scientists and ML Ops teams to collaborate and bring ML to market faster
Data Wrangler
A faster, visual way to aggregate and
prepare data for machine learning
Canvas
A visual point-and-click interface that allows analysts to generate accurate ML predictions on their own — without
requiring any machine learning experience or having to write a single line of code.
A dedicated no-code workspace for data analysts to generate ML-powered predictions
Amazon SageMaker Studio
Amazon SageMaker Canvas
Data
Science
Teams
Business
Teams
+
Many
deployment
options
Collaboration
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon SageMaker JumpStart ML hub
Share ML artifacts and access external hubs from Hugging Face, TensorFlow, and PyTorch
• Share ML models and notebooks with
other users in your AWS account
• Add ML artifacts developed with
SageMaker as well as those
developed outside of SageMaker
• Provides access to popular hubs
including Hugging Face, TensorFlow
Hub, and PyTorch Hub
• Browse and select shared models to
fine-tune, deploy endpoints, or run
notebooks
• Access foundational models with
billions of parameters
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
SageMaker JumpStart foundational models
Billions of parameters and adaptable to many use cases for NLP and computer vision
AlexaTM 20B model
• Alexa Teacher Model (AlexaTM) builds large-scale, multi-task, multi-lingual models
Stable Diffusion by StabilityAI
• Stable Diffusion generates images from given text
Bloom by Hugging Face
• Bloom models complete sentences or generate long paragraphs (46 languages)
Jurassic by AI21 Labs
• Apply to virtually any language task by giving a description and a few examples
pre:Invent
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
New Amazon SageMaker Canvas features
pre:Invent
Correlation matrices for advanced data analysis
• Correlation matrices allow you to summarize a dataset into a matrix that shows
correlations between two or more values and how they relate to one another
• Helps to identify and visualize patterns in a given dataset for advanced analysis
Encryption support with customer managed keys for time series
forecast models
• SageMaker Canvas now supports encryption at rest using CMK with AWS KMS for all
problem types currently supported by SageMaker Canvas.
Tags to track and allocate costs incurred by users
• Assign tags to user-profiles created within Amazon SageMaker to track SageMaker
Canvas usage costs categorized by users, departments, lines of businesses, or cost
centers
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
New Amazon SageMaker Canvas features
Bring your own models (BYOM) into SageMaker Canvas to generate predictions
Bring Your Own Model (BYOM) into
SageMaker Canvas
• Register model in SageMaker Model
Registry, and share
• Share models directly from
SageMaker Autopilot and JumpStart
Improved Model Sharing and
Collaboration from SageMaker Canvas
with SageMaker Studio Users
• Share models built in SageMaker
Canvas with data scientists using
SageMaker Studio for review, update,
and feedback
Dec 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
New Amazon SageMaker Autopilot features
pre:Invent
Ensemble training mode powered by AutoGluon
• Runs multiple trials with different combinations of a subset of algorithms and
AutoGluon configuration parameters
• Uses both the best objective metrics and lowest inference latency to select the best
model candidate for an experiment
Batch inference in Amazon SageMaker Studio
• Select any of the SageMaker Autopilot models and proceed with batch inference
within SageMaker Studio
AutoML experiments are now up to 2x faster in HPO training mode
• New multi-fidelity hyperparameter optimization (HPO) strategy employs state-of-
the-art hyperband tuning algorithm on datasets that are greater than 100 MB with
100 or more trials
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
AutoML support in Amazon SageMaker Pipelines
Launch Amazon SageMaker Autopilot jobs from SageMaker Pipelines
• SageMaker Autopilot is now integrated
with SageMaker Pipelines for automated
machine learning
• Add an automated training step
(AutoMLStep) in SageMaker Pipelines
and invoke a SageMaker Autopilot
experiment with Ensemble training
mode
Amazon SageMaker
Pipelines
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AI services
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
AWS AI Service Cards
• A new resource for Responsible AI
• Documents expected use cases,
limitations, design guidelines for
Responsible AI, and best practices
for use and operation
• Available today: Rekognition Face
Matching, Textract AnalyzeID, and
Transcribe Batch (English-US)
• Will be expanded based on customer
feedback
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Textract
pre:Invent
Updates to the Analyze ID API
• Data extraction for the machine readable zone, or MRZ code, on U.S. Passports.
Updates to the Analyze Expense API
• Support for 40+ normalized fields, including both summary fields such as Vendor
Address, and line item fields such as Product Code.
Updates to the forms and text extraction features
• Enhanced key-value pair extraction accuracy for single character boxed forms
commonly found in documents, such as Tax and Immigration forms.
• E13B fonts support commonly found in deposit checks/cheques
Detect signatures on any document
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Textract – Analyze Lending API
Amazon
Textract
Payslip
Identity
document
Bank
Statement
Extracted
Data
User
Review
Automated
Review
Approve
Reject
• Analyze and classify documents
contained in mortgage loan
applications
• Greater workflow automation to
accelerate automation efforts
• Reduce human error so that
users can focus on higher-value
tasks
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Transcribe – Real-Time Call Analytics
• Assist contact center agents in
resolving live calls faster
• Transcribe live calls, identify
customer experience issues and
sentiment in real time
• Combines automatic speech
NLP models trained to improve
overall customer experience
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Amazon Comprehend for IDP
Intelligent Document Processing
Amazon Comprehend
PDF
Microsoft
Word
Images
• Classify and extract entities
from files, without extracting
the text first
• Real-time inferencing of files,
as well as asynchronous batch
processing on large document
sets
• Combines OCR and
Comprehend NLP capabilities
to classify and extract entities
© 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved.
GA
Tabular search for HTML documents
Search more intuitively and effectively through tables
embedded in HTML pages
Extended language support for
semantic search
Kendra now supports semantic search for English,
Spanish, French, German, Portuguese, Japanese,
Korean, and Chinese
Credit Card Interest Rates
Bank 1 21.55
Bank 2 20.45
Bank 3 21.47
What’s the credit card with the lowest annual fees?
Credit Card Interest Rates
Bank 1 21.55
Bank 2 20.45
Bank 3 21.47
¿Qué es Amazon Kendra?
Qu'est-ce que Amazon Kendra ?
Was ist Amazon Kendra?
O que é a Amazon Kendra?
アマゾンケンドラとは?
Amazon Kendra란 무엇입니까?
什么是 Amazon Kendra?
什麼是 Amazon Kendra?
Amazon Kendra
Intelligent Enterprise Search
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
ML education
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
New fairness and bias mitigation course
• Free, public course on fairness
criteria and bias mitigation
• Taught by Amazon data scientists
who train AWS employees on ML
• 9+ hours of lectures
and hands-on exercises
>>> Get started today
Machine Learning
University
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Free educator enablement program
• AI & ML educator training program
for community colleges and MSIs
nationwide
• Hands-on training sessions
• Structured curriculum
and classroom resources
• Access to an educator
community of practice
>>> Learn more
Machine Learning
University
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
New 3-day course for Amazon SageMaker
New 3-day classroom course available
What you’ll learn:
• Accelerate the preparation,
building, training,
deployment, and monitoring
of ML solutions by using
Amazon SageMaker Studio
• Use the tools that are part of
SageMaker Studio to
improve productivity at
every step of the ML lifecycle
>>> Find a class today
Dec 2022
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Thank you!
Chris Fregly
Principal Specialist Solution Architect @
AWS AI/ML

More Related Content

What's hot

Using AWS Control Tower to govern multi-account AWS environments at scale - G...
Using AWS Control Tower to govern multi-account AWS environments at scale - G...Using AWS Control Tower to govern multi-account AWS environments at scale - G...
Using AWS Control Tower to govern multi-account AWS environments at scale - G...
Amazon Web Services
 

What's hot (20)

Intro to SageMaker
Intro to SageMakerIntro to SageMaker
Intro to SageMaker
 
AWS Data Analytics on AWS
AWS Data Analytics on AWSAWS Data Analytics on AWS
AWS Data Analytics on AWS
 
An Overview of Machine Learning on AWS
An Overview of Machine Learning on AWSAn Overview of Machine Learning on AWS
An Overview of Machine Learning on AWS
 
Executing a Large-Scale Migration to AWS
Executing a Large-Scale Migration to AWSExecuting a Large-Scale Migration to AWS
Executing a Large-Scale Migration to AWS
 
Deploy and Govern at Scale with AWS Control Tower
Deploy and Govern at Scale with AWS Control TowerDeploy and Govern at Scale with AWS Control Tower
Deploy and Govern at Scale with AWS Control Tower
 
Data & Analytics ReInvent Recap [AWS Basel Meetup - Jan 2023].pdf
Data & Analytics ReInvent Recap [AWS Basel Meetup - Jan 2023].pdfData & Analytics ReInvent Recap [AWS Basel Meetup - Jan 2023].pdf
Data & Analytics ReInvent Recap [AWS Basel Meetup - Jan 2023].pdf
 
How to Streamline DataOps on AWS
How to Streamline DataOps on AWSHow to Streamline DataOps on AWS
How to Streamline DataOps on AWS
 
Using AWS Control Tower to govern multi-account AWS environments at scale - G...
Using AWS Control Tower to govern multi-account AWS environments at scale - G...Using AWS Control Tower to govern multi-account AWS environments at scale - G...
Using AWS Control Tower to govern multi-account AWS environments at scale - G...
 
Simplify & Standardise your migration to AWS with a Migration Landing Zone
Simplify & Standardise your migration to AWS with a Migration Landing ZoneSimplify & Standardise your migration to AWS with a Migration Landing Zone
Simplify & Standardise your migration to AWS with a Migration Landing Zone
 
AWS Lake Formation Deep Dive
AWS Lake Formation Deep DiveAWS Lake Formation Deep Dive
AWS Lake Formation Deep Dive
 
Introduction to AWS Lake Formation.pptx
Introduction to AWS Lake Formation.pptxIntroduction to AWS Lake Formation.pptx
Introduction to AWS Lake Formation.pptx
 
Big Data Analytics Architectural Patterns and Best Practices (ANT201-R1) - AW...
Big Data Analytics Architectural Patterns and Best Practices (ANT201-R1) - AW...Big Data Analytics Architectural Patterns and Best Practices (ANT201-R1) - AW...
Big Data Analytics Architectural Patterns and Best Practices (ANT201-R1) - AW...
 
From Insights to Action, How to build and maintain a Data Driven Organization...
From Insights to Action, How to build and maintain a Data Driven Organization...From Insights to Action, How to build and maintain a Data Driven Organization...
From Insights to Action, How to build and maintain a Data Driven Organization...
 
Generative AI for the rest of us
Generative AI for the rest of usGenerative AI for the rest of us
Generative AI for the rest of us
 
Best Practice on using Azure OpenAI Service
Best Practice on using Azure OpenAI ServiceBest Practice on using Azure OpenAI Service
Best Practice on using Azure OpenAI Service
 
Automated Solution for Deploying AWS Landing Zone (GPSWS407) - AWS re:Invent ...
Automated Solution for Deploying AWS Landing Zone (GPSWS407) - AWS re:Invent ...Automated Solution for Deploying AWS Landing Zone (GPSWS407) - AWS re:Invent ...
Automated Solution for Deploying AWS Landing Zone (GPSWS407) - AWS re:Invent ...
 
Exploring Opportunities in the Generative AI Value Chain.pdf
Exploring Opportunities in the Generative AI Value Chain.pdfExploring Opportunities in the Generative AI Value Chain.pdf
Exploring Opportunities in the Generative AI Value Chain.pdf
 
Enterprise Governance: Build Your AWS Landing Zone (ENT351-R1) - AWS re:Inven...
Enterprise Governance: Build Your AWS Landing Zone (ENT351-R1) - AWS re:Inven...Enterprise Governance: Build Your AWS Landing Zone (ENT351-R1) - AWS re:Inven...
Enterprise Governance: Build Your AWS Landing Zone (ENT351-R1) - AWS re:Inven...
 
Building a Modern Data Architecture on AWS - Webinar
Building a Modern Data Architecture on AWS - WebinarBuilding a Modern Data Architecture on AWS - Webinar
Building a Modern Data Architecture on AWS - Webinar
 
How to go from zero to data lakes in days - ADB202 - New York AWS Summit
How to go from zero to data lakes in days - ADB202 - New York AWS SummitHow to go from zero to data lakes in days - ADB202 - New York AWS Summit
How to go from zero to data lakes in days - ADB202 - New York AWS Summit
 

Similar to AWS reInvent 2022 reCap AI/ML and Data

在 AWS 上構建無服務器分析
在 AWS 上構建無服務器分析在 AWS 上構建無服務器分析
在 AWS 上構建無服務器分析
Amazon Web Services
 

Similar to AWS reInvent 2022 reCap AI/ML and Data (20)

Single View of Data
Single View of DataSingle View of Data
Single View of Data
 
Re:cap día 1 del Aws Re:Invent 2023 - AWS UG Chile
Re:cap día 1 del Aws Re:Invent 2023 - AWS UG ChileRe:cap día 1 del Aws Re:Invent 2023 - AWS UG Chile
Re:cap día 1 del Aws Re:Invent 2023 - AWS UG Chile
 
Get More from your Data: Accelerate Time-to-Value and Reduce TCO with Conflue...
Get More from your Data: Accelerate Time-to-Value and Reduce TCO with Conflue...Get More from your Data: Accelerate Time-to-Value and Reduce TCO with Conflue...
Get More from your Data: Accelerate Time-to-Value and Reduce TCO with Conflue...
 
Building Modern Streaming Analytics with Confluent on AWS
Building Modern Streaming Analytics with Confluent on AWSBuilding Modern Streaming Analytics with Confluent on AWS
Building Modern Streaming Analytics with Confluent on AWS
 
AWS re-Invent re-Cap general deck 2022-2023 .pdf
AWS re-Invent re-Cap general deck 2022-2023 .pdfAWS re-Invent re-Cap general deck 2022-2023 .pdf
AWS re-Invent re-Cap general deck 2022-2023 .pdf
 
[Keynote] Accelerating Business Outcomes with AWS Data - 발표자: Saeed Gharadagh...
[Keynote] Accelerating Business Outcomes with AWS Data - 발표자: Saeed Gharadagh...[Keynote] Accelerating Business Outcomes with AWS Data - 발표자: Saeed Gharadagh...
[Keynote] Accelerating Business Outcomes with AWS Data - 발표자: Saeed Gharadagh...
 
Best Practices to Secure Data Lake on AWS (ANT327) - AWS re:Invent 2018
Best Practices to Secure Data Lake on AWS (ANT327) - AWS re:Invent 2018Best Practices to Secure Data Lake on AWS (ANT327) - AWS re:Invent 2018
Best Practices to Secure Data Lake on AWS (ANT327) - AWS re:Invent 2018
 
PaaS or Fail: Rule the Cloud with Altus
PaaS or Fail: Rule the Cloud with AltusPaaS or Fail: Rule the Cloud with Altus
PaaS or Fail: Rule the Cloud with Altus
 
Train Models on Amazon SageMaker Using Data Not from Amazon S3 (AIM419) - AWS...
Train Models on Amazon SageMaker Using Data Not from Amazon S3 (AIM419) - AWS...Train Models on Amazon SageMaker Using Data Not from Amazon S3 (AIM419) - AWS...
Train Models on Amazon SageMaker Using Data Not from Amazon S3 (AIM419) - AWS...
 
Better Together: Delivering Graph Value with AWS & Neo4j - Antony Prasad The...
Better Together:  Delivering Graph Value with AWS & Neo4j - Antony Prasad The...Better Together:  Delivering Graph Value with AWS & Neo4j - Antony Prasad The...
Better Together: Delivering Graph Value with AWS & Neo4j - Antony Prasad The...
 
Module 1 - CP Datalake on AWS
Module 1 - CP Datalake on AWSModule 1 - CP Datalake on AWS
Module 1 - CP Datalake on AWS
 
PaaS or Fail: Rule the Cloud with Altus
PaaS or Fail: Rule the Cloud with AltusPaaS or Fail: Rule the Cloud with Altus
PaaS or Fail: Rule the Cloud with Altus
 
AWS Lambda Powertools walkthrough.pdf
AWS Lambda Powertools walkthrough.pdfAWS Lambda Powertools walkthrough.pdf
AWS Lambda Powertools walkthrough.pdf
 
Build Data Lakes and Analytics on AWS: Patterns & Best Practices
Build Data Lakes and Analytics on AWS: Patterns & Best PracticesBuild Data Lakes and Analytics on AWS: Patterns & Best Practices
Build Data Lakes and Analytics on AWS: Patterns & Best Practices
 
Build Data Lakes & Analytics on AWS: Patterns & Best Practices
Build Data Lakes & Analytics on AWS: Patterns & Best PracticesBuild Data Lakes & Analytics on AWS: Patterns & Best Practices
Build Data Lakes & Analytics on AWS: Patterns & Best Practices
 
Managed Relational Databases
Managed Relational DatabasesManaged Relational Databases
Managed Relational Databases
 
Amazon EMR과 SageMaker를 이용하여 데이터를 준비하고 머신러닝 모델 개발 하기
Amazon EMR과 SageMaker를 이용하여 데이터를 준비하고 머신러닝 모델 개발 하기Amazon EMR과 SageMaker를 이용하여 데이터를 준비하고 머신러닝 모델 개발 하기
Amazon EMR과 SageMaker를 이용하여 데이터를 준비하고 머신러닝 모델 개발 하기
 
Module 3 - QuickSight Overview
Module 3 - QuickSight OverviewModule 3 - QuickSight Overview
Module 3 - QuickSight Overview
 
在 AWS 上構建無服務器分析
在 AWS 上構建無服務器分析在 AWS 上構建無服務器分析
在 AWS 上構建無服務器分析
 
Replicate & Manage Data Using Managed Databases & Serverless Technologies (DA...
Replicate & Manage Data Using Managed Databases & Serverless Technologies (DA...Replicate & Manage Data Using Managed Databases & Serverless Technologies (DA...
Replicate & Manage Data Using Managed Databases & Serverless Technologies (DA...
 

More from Chris Fregly

Amazon reInvent 2020 Recap: AI and Machine Learning
Amazon reInvent 2020 Recap:  AI and Machine LearningAmazon reInvent 2020 Recap:  AI and Machine Learning
Amazon reInvent 2020 Recap: AI and Machine Learning
Chris Fregly
 
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
Chris Fregly
 
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
Chris Fregly
 
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
Chris Fregly
 
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
Chris Fregly
 
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
Chris Fregly
 
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
Chris Fregly
 

More from Chris Fregly (20)

Pandas on AWS - Let me count the ways.pdf
Pandas on AWS - Let me count the ways.pdfPandas on AWS - Let me count the ways.pdf
Pandas on AWS - Let me count the ways.pdf
 
Ray AI Runtime (AIR) on AWS - Data Science On AWS Meetup
Ray AI Runtime (AIR) on AWS - Data Science On AWS MeetupRay AI Runtime (AIR) on AWS - Data Science On AWS Meetup
Ray AI Runtime (AIR) on AWS - Data Science On AWS Meetup
 
Smokey and the Multi-Armed Bandit Featuring BERT Reynolds Updated
Smokey and the Multi-Armed Bandit Featuring BERT Reynolds UpdatedSmokey and the Multi-Armed Bandit Featuring BERT Reynolds Updated
Smokey and the Multi-Armed Bandit Featuring BERT Reynolds Updated
 
Amazon reInvent 2020 Recap: AI and Machine Learning
Amazon reInvent 2020 Recap:  AI and Machine LearningAmazon reInvent 2020 Recap:  AI and Machine Learning
Amazon reInvent 2020 Recap: AI and Machine Learning
 
Waking the Data Scientist at 2am: Detect Model Degradation on Production Mod...
Waking the Data Scientist at 2am:  Detect Model Degradation on Production Mod...Waking the Data Scientist at 2am:  Detect Model Degradation on Production Mod...
Waking the Data Scientist at 2am: Detect Model Degradation on Production Mod...
 
Quantum Computing with Amazon Braket
Quantum Computing with Amazon BraketQuantum Computing with Amazon Braket
Quantum Computing with Amazon Braket
 
15 Tips to Scale a Large AI/ML Workshop - Both Online and In-Person
15 Tips to Scale a Large AI/ML Workshop - Both Online and In-Person15 Tips to Scale a Large AI/ML Workshop - Both Online and In-Person
15 Tips to Scale a Large AI/ML Workshop - Both Online and In-Person
 
AWS Re:Invent 2019 Re:Cap
AWS Re:Invent 2019 Re:CapAWS Re:Invent 2019 Re:Cap
AWS Re:Invent 2019 Re:Cap
 
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
KubeFlow + GPU + Keras/TensorFlow 2.0 + TF Extended (TFX) + Kubernetes + PyTo...
 
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
Swift for TensorFlow - Tanmay Bakshi - Advanced Spark and TensorFlow Meetup -...
 
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
Hands-on Learning with KubeFlow + Keras/TensorFlow 2.0 + TF Extended (TFX) + ...
 
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
Spark SQL Catalyst Optimizer, Custom Expressions, UDFs - Advanced Spark and T...
 
PipelineAI Continuous Machine Learning and AI - Rework Deep Learning Summit -...
PipelineAI Continuous Machine Learning and AI - Rework Deep Learning Summit -...PipelineAI Continuous Machine Learning and AI - Rework Deep Learning Summit -...
PipelineAI Continuous Machine Learning and AI - Rework Deep Learning Summit -...
 
PipelineAI Real-Time Machine Learning - Global Artificial Intelligence Confer...
PipelineAI Real-Time Machine Learning - Global Artificial Intelligence Confer...PipelineAI Real-Time Machine Learning - Global Artificial Intelligence Confer...
PipelineAI Real-Time Machine Learning - Global Artificial Intelligence Confer...
 
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
Hyper-Parameter Tuning Across the Entire AI Pipeline GPU Tech Conference San ...
 
PipelineAI Optimizes Your Enterprise AI Pipeline from Distributed Training to...
PipelineAI Optimizes Your Enterprise AI Pipeline from Distributed Training to...PipelineAI Optimizes Your Enterprise AI Pipeline from Distributed Training to...
PipelineAI Optimizes Your Enterprise AI Pipeline from Distributed Training to...
 
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
 
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
High Performance Distributed TensorFlow in Production with GPUs - NIPS 2017 -...
 
PipelineAI + TensorFlow AI + Spark ML + Kuberenetes + Istio + AWS SageMaker +...
PipelineAI + TensorFlow AI + Spark ML + Kuberenetes + Istio + AWS SageMaker +...PipelineAI + TensorFlow AI + Spark ML + Kuberenetes + Istio + AWS SageMaker +...
PipelineAI + TensorFlow AI + Spark ML + Kuberenetes + Istio + AWS SageMaker +...
 
PipelineAI + AWS SageMaker + Distributed TensorFlow + AI Model Training and S...
PipelineAI + AWS SageMaker + Distributed TensorFlow + AI Model Training and S...PipelineAI + AWS SageMaker + Distributed TensorFlow + AI Model Training and S...
PipelineAI + AWS SageMaker + Distributed TensorFlow + AI Model Training and S...
 

Recently uploaded

Recently uploaded (20)

Tech Trends Report 2024 Future Today Institute.pdf
Tech Trends Report 2024 Future Today Institute.pdfTech Trends Report 2024 Future Today Institute.pdf
Tech Trends Report 2024 Future Today Institute.pdf
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemkeProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...2024: Domino Containers - The Next Step. News from the Domino Container commu...
2024: Domino Containers - The Next Step. News from the Domino Container commu...
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your Business
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 

AWS reInvent 2022 reCap AI/ML and Data

  • 1. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. 2022 AI/ML and Data Edition Chris Fregly Principal Specialist Solution Architect @ AWS AI/ML
  • 2. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. A P P S D E V I C E S P E O P L E A P P / L O G S T H I R D - P A R T Y D A T A I O T / D E V I C E S Data sources F O R A P P L I C A T I O N S Amazon Aurora Amazon Kinesis & Amazon MSK F O R A N A L Y T I C S A N D M A C H I N E L E A R N I N G Data Lake Amazon S3 Amazon Redshift Data Warehouse Amazon Redshift Amazon EMR B U S I N E S S I N T E L L I G E N C E Amazon QuickSight M A C H I N E L E A R N I N G Amazon SageMaker A N A L Y T I C S Amazon DynamoDB AWS Glue | AWS Lake Formation, Amazon DataZone Building an end-to-end ML and data strategy
  • 3. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Agenda ML infrastructure and hardware ML and data governance Discover, analyze, and prepare data Build and train ML models Deploy ML models for inference Low-code / no-code ML AI services ML education
  • 4. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. ML infrastructure and hardware
  • 5. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Journey of silicon innovation at AWS AWS Inferentia and AWS Trainium Machine learning acceleration AWS Graviton Powerful and efficient, modern applications AWS Nitro System Hypervisor, Nitro Cards, network, storage, SSD, and security
  • 6. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker supports new instance types SageMaker Model Training support for ml.trn1 instances • Powered by AWS Trainium chips • ml.trn1.2xlarge, for experimenting with a single accelerator and training small models cost effectively • ml.trn1.32xlarge for training large-scale models SageMaker Inference adds eight new Graviton-based instances • Powered by Graviton3 and Graviton2 • For Real-time and asynchronous inference model deployment options • Graviton3: ml.c7g • Graviton2: ml.m6g, ml.m6gd, ml.c6g, ml.c6gd, ml.c6gn, ml.r6g, and ml.r6gd Trn1 C7g M6g C6g R6g p r e : I n v e n t
  • 7. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview New instance types for Amazon SageMaker P4de instance documentation P4de • Provide the highest performance for ML training and HPC applications • Powered by 8 NVIDIA A100 GPUs with 80 GB high-performance HBM2e GPU memory, 2X higher than the GPUs in current P4d instances • Up to 640GB of GPU memory, providing up to 60 percent better ML training performance along with 20 percent lower cost to train when compared to P4d instances SageMaker Model Training support for ml.p4de.24xlarge instances Dec 2022
  • 8. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. ML and data governance
  • 9. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Why ML and data governance? 9 Machine learning (ML) governance Onboard Develop Monitor User setup ML activities Deployment Build Train Prepare Tune Inferences Customers Business applications Platform admin Data engineer Data scientist ML engineer ML risk officer Model approver
  • 10. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker – New ML Governance Tools Simplify access control and enhance transparency Amazon SageMaker Role Manager Define custom user permissions in minutes Amazon SageMaker Model Cards Centralize model information and documentation Amazon SageMaker Model Dashboard Monitor model performance in one place
  • 11. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Role Manager Define custom user permissions in minutes • Simplify permissions for ML activities • Use guided workflows for role creation • Accelerate user onboarding
  • 12. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Model Cards Easily document, retrieve, and share the necessary model information • Streamline model documentation • Capture model information, such as input datasets, training environments, training results, model purpose, performance goals • Attach and visualize evaluation results, such as bias and quality metrics • Share model cards with business stakeholders, internal teams, or your customers
  • 13. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Model Dashboard Unified view across all your models to audit performance • Track model behavior • Integrates with SageMaker Model Monitor and SageMaker Clarify • Monitor model behavior for data quality, model quality, bias drift, and feature attribution drift • Automate alerts • Troubleshoot model deviations
  • 14. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Coming Soon Share data securely with Amazon DataZone Unlock data across organizational boundaries with build-in governance Amazon DataZone Data producers Data consumers Fine-grained controls to manage and govern access to data Discover and share data at scale across organizational boundaries Makes it easy for data scientists and other business users to discover, use, and collaborate around that data Manage organization-wide governance in one place
  • 15. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA SageMaker Studio passes user permissions to EMR Multiple SageMaker Studio users can connect to the same EMR cluster with isolated access • All EMR jobs created from SageMaker Studio will inherit data and resources permissions for the given user. • Multiple SageMaker Studio users can use the same EMR cluster with separate access to data • When accessing data lakes managed by Lake Formation, table-level and column-level access policies are enforced p r e : I n v e n t
  • 16. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Dynamic data masking with Amazon Redshift Protect sensitive data with role-based permissions ID Geo- location Name Phone number 123 WA Ana 123-456-3568 124 NY Alice 123-457-**** 125 WA Bruce 123-457-3569 126 CA Chris 123-457-**** 130 CA Sharon 123-457-**** Condition column Mask column
  • 17. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Granular access with Redshift and Lake Formation Centrally manage data sharing in Amazon Redshift with AWS Lake Formation Amazon Redshift Amazon Redshift Amazon Redshift AWS Lake Formation
  • 18. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Amazon Security Lake Automatically centralize security data into a purpose-built data lake in a few steps
  • 19. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Discover, analyze, and prepare data
  • 20. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Integration with many popular data sources Supports both AWS and 3rd-party data source integrations from Amazon SageMaker • Amazon S3, Athena, Redshift, EMR • Salesforce, ServiceNow, Marketo • Mailchimp, SendGrid, Zendesk, Jira, Datadog
  • 21. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview AWS Data Exchange for AWS Lake Formation Secure data mesh architecture with third-party data using AWS Data Exchange AWS Lake Formation Providers use LF-tags to indicate which data subscribers should have access to Third-party data provider Provider stores data in AWS Lake Formation AWS Data Exchange Grants subscriber read- only access to the data tagged with the key-value pairs specified by the provider when a subscription starts and automatically denies access when it ends AWS Marketplace Automated payments and billing Subscriber Subscribers see the data once their account is verified Query provider data Query, transform, or share access with the appropriate user groups without any upfront ETL AWS account + AWS Data Exchange for AWS Lake Formation
  • 22. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Amazon CodeWhisperer code generator Enterprise administrative controls, simple sign-up, and support for new languages • Generates code recommendations based on the comments – and prior code - in your IDE • Available in popular IDEs such as Visual Studio Code, JetBrains, AWS Cloud9, AWS Lambda console • Supports Python, Java, JavaScript, C#, TypeScript • Enable CodeWhisperer for your organization with single sign-in (SSO) authentication • Sign-up with AWS Builder ID • Generates open source attribution documentation for you
  • 23. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview AWS Glue Data Quality continuous monitoring Deliver high quality data across your data lakes and data pipelines • Automatic data quality rule recommendations based on your data • Keep data quality high with ongoing data analysis and quality checking • Data quality for datasets in your data lake and data pipelines • Cost-effective to scale with pay-as-you-go billing, with no lock-in AWS Glue Data Quality Amazon SageMaker
  • 24. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. AWS Glue for Ray Scaling your data integration workloads using Python • AWS Glue for Ray is a new engine option on AWS Glue. • Data engineers and ML practitioners can use AWS Glue for Ray to process large datasets with Python and popular Python libraries. • AWS Glue jobs are fire-and-forget systems where you can submit your Ray code to the AWS Glue jobs API • AWS Glue for Ray facilitates the distributed processing of your Python code over multi- node clusters. AWS Glue Amazon SageMaker Preview
  • 25. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA • Access interactive Spark clusters that start in under a second and run faster with our optimized runtime for Apache Spark • Harness Apache Spark for complex, powerful analytics using the expressive power of Python along with its wide ecosystem • Build Apache Spark applications without managing resources or configuring software using Amazon Athena Amazon Athena Amazon Athena for Apache Spark Run interactive analytics on Apache Spark in under a second Amazon SageMaker
  • 26. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Redshift integration for Apache Spark Build Interactive Spark Applications with Amazon SageMaker, Glue, and EMR Redshift Connector for Apache Spark Amazon Redshift Amazon Glue Amazon EMR • Apache Spark applications accessing Amazon Redshift data from AWS analytics services such as Amazon EMR, AWS Glue, and Amazon SageMaker • Build Apache Spark applications that read from and write to your Amazon Redshift data warehouse, without compromising performance or transactional consistency. • No manual setup and maintenance of uncertified versions of Spark-Redshift open-source connectors • Improved performance with only relevant data moved from Redshift to consuming applications Amazon SageMaker
  • 27. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Streaming data support for Amazon Redshift Directly ingest streaming data into your data warehouse for real-time processing • Directly ingest streaming data into Amazon Redshift from Kinesis Data Streams and Managed Streaming for Apache Kafka without staging in S3 • Perform rich analytics using familiar SQL on streaming data • Easily create and manage extract-load-transform (ELT) pipelines with streaming data • Process large volumes of streaming data from multiple sources to derive insights in seconds Amazon SageMaker Amazon Kinesis Data Streams Amazon Managed Streaming for Apache Kafka Redshift Kinesis or Kafka producer KDS Stream MSK Topic
  • 28. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Multi-AZ support for Amazon Redshift Highly resistant data warehouse with auto-failover and no data loss • Workload processing across availability zones (AZs) • Easy management through a single endpoint • Auto-failover with zero data loss and no manual intervention Amazon Redshift managed storage AZ 1 AZ 2
  • 29. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Amazon Redshift Auto-Copy from Amazon S3 Simplified and automated file ingestion from Amazon S3 into Redshift • Simple, low code data ingestion • Avoid re-ingestion and manual tracking of loaded files • Easily convert your existing COPY statements into automatic ingestion jobs • Automatic ingestion of new data from Amazon S3 based on user defined configurations Amazon S3 Redshift Copy Job Redshift Table Continuously monitoring S3 folder New file(s) detected Ingestion automatically starts Amazon SageMaker
  • 30. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Zero-ETL integration from Redshift to Aurora Access multiple Amazon Aurora databases with Amazon Redshift • Drive holistic insights across applications or partitions • Analyze data from multiple Aurora databases in the same Redshift cluster • Leverage Redshift features such as materialized views, data sharing and federated access to data lakes Amazon Redshift Amazon Aurora Amazon SageMaker
  • 31. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Serverless EMR and Redshift are now GA Pay only for the resources you use with these serverless options Amazon EMR Serverless • Run Spark and Hive applications without having to configure, optimize, or manage clusters • Fine-grained auto-scaling of compute and memory resources • Uses the performance-optimized EMR runtime Amazon Redshift Serverless • Run analytics queries without having to configure, tune, and manage data warehouse clusters • Intelligently auto-scales data warehouse capacity to match your workload demand in seconds • Supports Redshift Query Editor v2 or any business intelligence (BI) tool of your choice Sum m er 2022
  • 32. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Data Wrangler new features Built-in data preparation in SageMaker Studio Notebooks • Automatically generates key visualizations on top of Pandas data frames • Understand data distribution and identify data quality issues • Generate ML-specific insights for ML target column • Receive recommendations for data transformations and code import pandas as pd import sagemaker_datawrangler df = pd.read_csv("data.csv")
  • 33. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Data Wrangler new features Deploy data preparation flows for real-time and batch inference Data Wrangler Flow Data Scientist ML Engineer Amazon SageMaker Data Wrangler Data Preparation Job Model Training Inference Pipeline Run data preparation for model training Reuse data transformation flow for real-time & batch inference Define data preparation for training Deploy inference • Deploy data preparation flows from SageMaker Data Wrangler for real-time and batch inference • Reuse the data transformation flow • Speed up your production deployment
  • 34. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker Data Wrangler + EMR/Athena Now supports Amazon EMR Presto and EMR Spark as big-data query engines • Connect to existing Amazon EMR Presto and Athena clusters using a visual experience in SageMaker Data Wrangler • Prepare data for ML in minutes using Data Wrangler’s visual interface • Analyze data, clean data, and create features for ML using 300+ built-in transformations backed by Spark without the need to author Spark code Amazon EMR Presto Amazon SageMaker Data Wrangler D e c 2 0 2 2 Amazon Athena
  • 35. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Apache Iceberg: Amazon SageMaker Feature Store Now supports Apache Iceberg table format • Create feature groups in the offline store in Apache Iceberg table format • Apache Iceberg is an open table format for very large analytic datasets • Apache Iceberg compacts small data files into fewer large files in the partition, resulting in significantly faster queries. Amazon SageMaker Feature Store Apache Iceberg D e c 2 0 2 2
  • 36. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Omics Store, query, analyze, and generate insights from genomic and other omics data • Built-in access control and logging D N A R N A P R O T E I N S Genomics Transcriptomics Proteomics
  • 37. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Expanded API capabilities for Amazon QuickSight Programmatic access to the underlying structure of QuickSight dashboards • Access underlying data models of Amazon QuickSight dashboards, reports, analyses and templates via the AWS Software Development Kit (SDK). • Translate legacy BI assets to cloud-native dashboards quickly • Reduce migration time from months to weeks • Integrate into DevOps processes such as code reviews, audits, and audit every change before deployment Amazon QuickSight
  • 38. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon QuickSight Paginated Reports Day-to-day detailed operational data in custom formats • Create, schedule, and share highly formatted multipage reports • Build all insights, independent of preferred consumption model, on single source of truth governed datasets • Single authoring experience for dashboards and reports • Unified consumption experience allows users to choose dashboards and reports • Pay for what you use. Amazon QuickSight
  • 39. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Build and train ML models
  • 40. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Amazon CodeWhisperer code generator Enterprise administrative controls, simple sign-up, and support for new languages • Generates code recommendations based on the comments – and prior code - in your IDE • Available in popular IDEs such as Visual Studio Code, JetBrains, AWS Cloud9, AWS Lambda console • Supports Python, Java, JavaScript, C#, TypeScript • Enable CodeWhisperer for your organization with single sign-in (SSO) authentication • Sign-up with AWS Builder ID • Generates open source attribution documentation for you
  • 41. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Improved Amazon SageMaker Studio usability Redesigned user interface (UI) and user experience (UX) based on customer feedback • Redesigned navigation menu following ML workflow • Dynamic landing pages with links to videos, tutorials, blogs, etc. • New SageMaker Home page with one-click access to common tasks • Redesigned launcher with quick links to create notebook, open console, etc. Supported for SageMaker Studio domains running on JupyterLab 3+
  • 42. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Notebook Jobs for Amazon SageMaker Studio Automatic conversion of notebook code to production-ready jobs • Run your notebooks as is or in a parameterized fashion with just a few simple clicks from the SageMaker Studio • Run notebooks on a schedule or immediately • No need for the end-user to modify their existing notebook code • View the populated notebook cells after the job is complete, including any visualizations • Also available in SageMaker Studio Lab
  • 43. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Improved Amazon SageMaker Experiments Next-generation experiment tracking, model comparison • New UI for easier experiment tracking and model comparison • Organize, track, compare and evaluate machine learning (ML) experiments and model versions from any IDE, including local Jupyter notebooks • Programmatic tracking of experiments and logging of metrics/parameters Dec 2022
  • 44. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Collaboration with Amazon SageMaker Studio Team-based sharing and real-time collaboration with shared spaces • SageMaker Studio now supports team-based sharing and real-time collaboration • Create shared spaces to access, read, edit, and share the same notebook in real time • Shared EFS directory in a space • Filtered SageMaker resources Experiments Model Registry Pipelines, etc.
  • 45. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Collaboration with Amazon SageMaker Studio Team-based sharing and real-time collaboration on the same notebooks
  • 46. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon SageMaker + Glue Interactive Sessions Scaling your data integration workloads using interactive Apache Spark and Ray • From SageMaker Studio, access AWS Glue Interactive Sessions to access large, serverless, distributed clusters for Apache Spark and Ray • Data engineers and ML practitioners can use Apache Spark or Ray to process large datasets with Python and popular Python libraries. • Train distributed ML models on large datasets using algorithms from Apache Spark or Ray AWS Glue Interactive Sessions Amazon SageMaker Preview
  • 47. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Heterogeneous clusters for Training Jobs Reduce training cost by using a mix of instance types specific to your workload • Use both CPU-optimized and GPU- optimized instance types in a single distributed training job • Perform data transformations on CPU instance types • Perform ML operations on GPU instance types O ct 2022
  • 48. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Warm pools for SageMaker Training Jobs Reduce startup time for SageMaker Training and Tuning Jobs with warm pools • Enables faster iterations for ML development lifecycle • Up to 8x improvement in startup times for SageMaker Training, Tuning, and Autopilot jobs • Balance cost and convenience with configurable keep-alive times Sept 2022 from sagemaker.pytorch import PyTorch estimator = PyTorch( entry_point='train.py’, ... keep_alive_period_in_seconds=1800) estimator.fit('s3://my_bucket/my_data/')
  • 49. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA SageMaker supports more domains and users Now supports multiple domains within the same AWS Account • Create multiple SageMaker domains within the same AWS account • Scope access and isolate resources to different team or business units in your organization
  • 50. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker tagging and cost-allocation Automated tagging for better cost allocation and monitoring • SageMaker automatically tags resources at domain, user and space level supporting detailed cost allocation • All SageMaker resources with an Amazon Resource Name (ARN) including: • Training jobs • Processing jobs • Kernel Gateways • Endpoints • etc. • Supports more detailed cost allocation for administrators
  • 51. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview • Bring your own or acquire geospatial data with just a few clicks • Easily prepare geospatial data with built-in operations and transformations • Speed model building with pre- trained deep neural network (DNN) models and geospatial operators • Analyze and explore predictions with built-in visualization tools Amazon SageMaker – Geospatial ML Build, train, and deploy ML models using geospatial data Aerial and satellite imagery Mapping data Road mask (color as speed) 24 Hour Fitness 7-Eleven 76 Gas Stations 84 Lumber 85C Bakery Cafe
  • 52. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Preview Amazon SageMaker – Geospatial ML Visualize predictions Select or train a model Access geospatial data sources
  • 53. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Updates to Amazon SageMaker Pipelines Supports local mode and cross-account artifact sharing Fully supports “local mode” for iterative development and testing • Fast feedback loops are critical to speed up pipeline development and testing • Develop and test pipelines locally to reduce cost and increase developer efficiency • Use on your laptop with any IDE such as VSCode, PyCharm, vi, or emacs Securely share pipeline artifacts across multiple accounts • Multi-account strategy helps achieve data, project, and team isolation • Share pipeline artifacts between lines of business within your organization • Supports different authorization levels including read-only and execute permissions Fall 2022
  • 54. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Deploy ML models for inference
  • 55. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Shadow testing for Amazon SageMaker Endpoints Now supports shadow testing to compare models in real-time • Validate performance of your models by comparing them to production models • SageMaker takes care of mirroring requests • Start small and dial up to control costs • Catch potential configuration errors and performance issues before they impact end users • Monitor progress of the shadow test and performance metrics such as latency and error rate through a live dashboard Amazon SageMaker Endpoint Production Variant Shadow Variant Model A Model B Request Response Request Request Response Application Response Amazon S3 Accessible through AWS console, CLI, APIs
  • 56. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA • Set up a test for a predefined duration • Monitor operational performance through a live dashboard • Deploy models with confidence or rollback Shadow testing for Amazon SageMaker Endpoints Tools to manage shadow tests
  • 57. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA More support for large models and GPUs Now supports deploying large models that require large volume size • Deploy large models (up to 500GB) for inference on SageMaker’s Real-time and Asynchronous Inference options by configuring the maximum EBS volume size and timeout quotas • Container health check and download timeout quotas have been made configurable up to 60 minutes, so you have more time to download and load your model and associated resources Multi Model Endpoint (MME) now supports GPU-based instances • Use MME to deploy thousands of ML models on GPU-based instances • MME dynamically loads and unloads models from GPU memory based on incoming traffic to the endpoint • Save cost with MME as the GPU instances are shared by thousands of models
  • 58. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Multi-model Endpoint (MME) support for GPUs Use MME to deploy thousands of ML models on GPU-based instances Run hundreds of models behind a single endpoint to maximize GPU utilization Powered by NVIDIA Triton Server to support run models trained on your choice of popular frameworks Wide selection of up to 15 GPU instance types Automatically applies endpoint autoscaling policy to all models P3 G4 G5 🤗 Hugging Face
  • 59. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Monitor Amazon SageMaker Batch Transform Jobs p r e : I n v e n t Now supported for Batch Transform Jobs in Amazon SageMaker • Monitor the quality of ML predictions from Batch Transform jobs in SageMaker using Amazon SageMaker Model Monitor • SageMaker Batch Transform enables you to run predictions on datasets stored in Amazon S3 • Collect Batch Transform data in production, analyze it, and compare it against your training or validation data to detect deviations • Use SageMaker Model Monitor’s built-in rules to detect drift for structured data sets, add data transformations before you run the built-in rules, or write your own custom rules
  • 60. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Low-code / no-code ML
  • 61. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon SageMaker low-code / no-code Business Requirements Data Preparation & Feature Engineering Model Development, Training, and Tuning Model Deployment Inference & Monitoring Autopilot AutoML capability that automatically prepares your data, as well as builds, trains, and tunes the best machine learning models for your tabular and time-series datasets JumpStart Pre-built solutions and a model zoo of pre-trained and easily tunable state-of-the-art models for Computer Vision, and Natural Language Processing A dedicated workspace for data engineers, data scientists and ML Ops teams to collaborate and bring ML to market faster Data Wrangler A faster, visual way to aggregate and prepare data for machine learning Canvas A visual point-and-click interface that allows analysts to generate accurate ML predictions on their own — without requiring any machine learning experience or having to write a single line of code. A dedicated no-code workspace for data analysts to generate ML-powered predictions Amazon SageMaker Studio Amazon SageMaker Canvas Data Science Teams Business Teams + Many deployment options Collaboration
  • 62. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon SageMaker JumpStart ML hub Share ML artifacts and access external hubs from Hugging Face, TensorFlow, and PyTorch • Share ML models and notebooks with other users in your AWS account • Add ML artifacts developed with SageMaker as well as those developed outside of SageMaker • Provides access to popular hubs including Hugging Face, TensorFlow Hub, and PyTorch Hub • Browse and select shared models to fine-tune, deploy endpoints, or run notebooks • Access foundational models with billions of parameters
  • 63. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA SageMaker JumpStart foundational models Billions of parameters and adaptable to many use cases for NLP and computer vision AlexaTM 20B model • Alexa Teacher Model (AlexaTM) builds large-scale, multi-task, multi-lingual models Stable Diffusion by StabilityAI • Stable Diffusion generates images from given text Bloom by Hugging Face • Bloom models complete sentences or generate long paragraphs (46 languages) Jurassic by AI21 Labs • Apply to virtually any language task by giving a description and a few examples pre:Invent
  • 64. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA New Amazon SageMaker Canvas features pre:Invent Correlation matrices for advanced data analysis • Correlation matrices allow you to summarize a dataset into a matrix that shows correlations between two or more values and how they relate to one another • Helps to identify and visualize patterns in a given dataset for advanced analysis Encryption support with customer managed keys for time series forecast models • SageMaker Canvas now supports encryption at rest using CMK with AWS KMS for all problem types currently supported by SageMaker Canvas. Tags to track and allocate costs incurred by users • Assign tags to user-profiles created within Amazon SageMaker to track SageMaker Canvas usage costs categorized by users, departments, lines of businesses, or cost centers
  • 65. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA New Amazon SageMaker Canvas features Bring your own models (BYOM) into SageMaker Canvas to generate predictions Bring Your Own Model (BYOM) into SageMaker Canvas • Register model in SageMaker Model Registry, and share • Share models directly from SageMaker Autopilot and JumpStart Improved Model Sharing and Collaboration from SageMaker Canvas with SageMaker Studio Users • Share models built in SageMaker Canvas with data scientists using SageMaker Studio for review, update, and feedback Dec 2022
  • 66. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA New Amazon SageMaker Autopilot features pre:Invent Ensemble training mode powered by AutoGluon • Runs multiple trials with different combinations of a subset of algorithms and AutoGluon configuration parameters • Uses both the best objective metrics and lowest inference latency to select the best model candidate for an experiment Batch inference in Amazon SageMaker Studio • Select any of the SageMaker Autopilot models and proceed with batch inference within SageMaker Studio AutoML experiments are now up to 2x faster in HPO training mode • New multi-fidelity hyperparameter optimization (HPO) strategy employs state-of- the-art hyperband tuning algorithm on datasets that are greater than 100 MB with 100 or more trials
  • 67. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. GA AutoML support in Amazon SageMaker Pipelines Launch Amazon SageMaker Autopilot jobs from SageMaker Pipelines • SageMaker Autopilot is now integrated with SageMaker Pipelines for automated machine learning • Add an automated training step (AutoMLStep) in SageMaker Pipelines and invoke a SageMaker Autopilot experiment with Ensemble training mode Amazon SageMaker Pipelines
  • 68. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. AI services
  • 69. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA AWS AI Service Cards • A new resource for Responsible AI • Documents expected use cases, limitations, design guidelines for Responsible AI, and best practices for use and operation • Available today: Rekognition Face Matching, Textract AnalyzeID, and Transcribe Batch (English-US) • Will be expanded based on customer feedback
  • 70. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Textract pre:Invent Updates to the Analyze ID API • Data extraction for the machine readable zone, or MRZ code, on U.S. Passports. Updates to the Analyze Expense API • Support for 40+ normalized fields, including both summary fields such as Vendor Address, and line item fields such as Product Code. Updates to the forms and text extraction features • Enhanced key-value pair extraction accuracy for single character boxed forms commonly found in documents, such as Tax and Immigration forms. • E13B fonts support commonly found in deposit checks/cheques Detect signatures on any document
  • 71. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Textract – Analyze Lending API Amazon Textract Payslip Identity document Bank Statement Extracted Data User Review Automated Review Approve Reject • Analyze and classify documents contained in mortgage loan applications • Greater workflow automation to accelerate automation efforts • Reduce human error so that users can focus on higher-value tasks
  • 72. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Transcribe – Real-Time Call Analytics • Assist contact center agents in resolving live calls faster • Transcribe live calls, identify customer experience issues and sentiment in real time • Combines automatic speech NLP models trained to improve overall customer experience
  • 73. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Amazon Comprehend for IDP Intelligent Document Processing Amazon Comprehend PDF Microsoft Word Images • Classify and extract entities from files, without extracting the text first • Real-time inferencing of files, as well as asynchronous batch processing on large document sets • Combines OCR and Comprehend NLP capabilities to classify and extract entities
  • 74. © 2023 Amazon Web Services, Inc. or its affiliates. All rights reserved. GA Tabular search for HTML documents Search more intuitively and effectively through tables embedded in HTML pages Extended language support for semantic search Kendra now supports semantic search for English, Spanish, French, German, Portuguese, Japanese, Korean, and Chinese Credit Card Interest Rates Bank 1 21.55 Bank 2 20.45 Bank 3 21.47 What’s the credit card with the lowest annual fees? Credit Card Interest Rates Bank 1 21.55 Bank 2 20.45 Bank 3 21.47 ¿Qué es Amazon Kendra? Qu'est-ce que Amazon Kendra ? Was ist Amazon Kendra? O que é a Amazon Kendra? アマゾンケンドラとは? Amazon Kendra란 무엇입니까? 什么是 Amazon Kendra? 什麼是 Amazon Kendra? Amazon Kendra Intelligent Enterprise Search
  • 75. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. ML education
  • 76. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. New fairness and bias mitigation course • Free, public course on fairness criteria and bias mitigation • Taught by Amazon data scientists who train AWS employees on ML • 9+ hours of lectures and hands-on exercises >>> Get started today Machine Learning University
  • 77. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Free educator enablement program • AI & ML educator training program for community colleges and MSIs nationwide • Hands-on training sessions • Structured curriculum and classroom resources • Access to an educator community of practice >>> Learn more Machine Learning University
  • 78. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. New 3-day course for Amazon SageMaker New 3-day classroom course available What you’ll learn: • Accelerate the preparation, building, training, deployment, and monitoring of ML solutions by using Amazon SageMaker Studio • Use the tools that are part of SageMaker Studio to improve productivity at every step of the ML lifecycle >>> Find a class today Dec 2022
  • 79. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. Thank you! Chris Fregly Principal Specialist Solution Architect @ AWS AI/ML