his will be a session from a technical and from a product and project management perspective. Building AI and ML into products requires a different mindset due to research required which requires engineers, product and project managers to have a unique process different from traditional product development. Further this can be complicated in a Multi tenant cloud environment to do it right. This session will provide some general practices that can help application engineers and data science engineers and project and product managers build in ai and ml in products.
Unlocking the Power of ChatGPT and AI in Testing - A Real-World Look, present...
Building Artificial Intelligence (AI) and Machine Learning products - Some General best practices
1. Building AI and ML into
products in a Multi tenant
cloud environment
Some General Best Practices –
Technology, Product, People, Project
Debashis Banerjee
(https://www.linkedin.com/in/debashisb/)
24th April 2019
Image Credit: Dreamscape
Disclaimer: All views expressed in this session are those of the author and does not represent those
of the organization he currently works or has worked in the past for
2. KEY QUESTIONS BEFORE DECIDING
TO BUILD AN AI/ML PRODUCT
• Do we see it as high volume and repeating?
• Do we have Accessible Data ?
• Do we have the Algorithms?
• Do we have access to scalable Storage and computer
power
• Do we have access to Skills & Learning mechanisms?
3. THE 5 STEP PROCESSTO BUILDING AN AI/ML BASED
PRODUCT ONTHE CLOUD
Build the UI
Lock your
storage
Connect the APIs
Productionize the
models
Demo, Find early
customers
Add Non
functional
Run a full project
Data
Analysis combined
with business
analysis
Data/Information
Design
Refine the Req
&
Don’t rush into
“project
management”
API Prototypes
Create re-usables
Step 1 Step 2 Step 3 Step 4 Step 5
Find the right
“problems”
Build the skills
Build the mindset
Find the
sponsors
Monitor
Data
Launch ,
Monitor, Scale
Repeat
Cross Train
Monitor the
costs
Re-skill
affected if
needed
Enable
Maintainers of
the Model
Model the costs
Find the skills
Structure team
right
One-Page Req
Talk to Legal
customer/user
segment
Detractors?
Find the data
Pick your GTM
4. FEW UNIQUE TECHNOLOGY ELEMENTSTO
AI/ML PRODUCTS
• MultiTenancy & avoiding co-mingling
• Anonymization
• Data security , GDPR
• Removing AI Bias
• Data set Diversity
• Should there be a version of Asimov’s Laws of
Robotics coded in?
• What design principles and tech to humanizing
AI/ML
Picture Credit:clearpathrobotics.com blog
5. SOME METRICS CATEGORIES
• Computing, Storage, External Data Source Recurring Cost
• Measure Complexity – hide or show, clicks, time to find, rate of
abandonment. customization
• Customer switch over cost / Training Cost
• Measure what does the AI/ML make Faster , Better, Cheaper
• Trial Availability – extent to which customer can try and then
adopt, Usage Metrics
• Pricing – explicit or implicit
• Innovation Diffusion
Innovation Diffusion:Source:Wikipedia
7. 5 PILLARS TO BUILDING AN AI/ML – “ME”
Technology Mindset Training Part Time
Industry
Bodies
8. 5 PILLARS TO BUILDING AN AI/ML – “WE”
(ORG)
Learning on
Data Science
topics
Temporary
Rotations
“Me-time”
PoCs
Hackathons Access to
Data
9. IN CONCLUSION – PULLING THE
“RABBIT” OUT OFTHE HAT
- Start with 5 pillars of becoming theAI/ML “Me”
- Support with with 5 pillars of being an AI/ML
“We”
- Pick the right problem
- Follow the 5 step process to build in AI/ML into
your cloud products
- Set up the product team right
- Understand the uniqueTechnology and
monitor theTechnology and business metrics