21. Expected benefits:
▪ Simplify our footprint on
tools’ variations
▪ Serverless, less things to
worry about
▪ Less operations overhead
▪ Scale easily
▪ Cheaper
WHY USING GOOGLE?
21
Main challenges:
▪ Different animals
▪ Compatibility with our
current workflow
▪ Might redo a lot of things,
especially reports and
enrichments
▪ Huge migrations
22. Tech @ Traveloka
Deployment Workflow
Code + PR
+
Automated
Test + CR
Manual Trigger
Deploy for Prod
(canary release,
stable release)
Automatic
Deploy to
Stg
24. Tech @ Traveloka
Data Use Case
Realperson: are you real?
Automate Know Your Customer
(KYC) process by leverage
off-the-shelf ML APIs such as
Google Cloud Vision and Azure
Face API.
ml-common: expose in-house ML model as web-service (in
collaboration with NVS team).
Flask-based web-service, with support for ML models from:
- caffe2 , Scikit Learn, Tensorflow Serving, Keras (w/ Tensorflow
Backend)
Hotel Images:
Automate hotel images
selection which improves the
hotel booking and transactions