Decision Making Based on Machine Learning at Outfittery
summary
Outfittery builds a production smart gateway that serves live A/B tests and process steering as REST-deployed deciders packed in Docker and orchestrated with Kubernetes, backed by Redis, SQL, RabbitMQ and real-time enrichment. Traffic sampling, experiment groups and a shadow mode let teams validate new algorithms before rollout; a two-person team ships what ten data scientists feed and a 25-person IT org operates. The concrete payoff is replacing eight-hour stale ETL with synchronous enrichments and fast, measurable revenue-impact experiments.
AI-written summary · about
chapters (13)
- 0:00Intro
- 3:05The problem
- 7:45Optimizely
- 10:05Tracking Conversion
- 12:17Smart Gateway
- 14:29Algorithms
- 16:00Configuration
- 16:47Sampling
- 21:54Making it better
- 23:28Changes
- 27:34Life is Good
- 30:46Service Computing
- 36:05Conclusion
related talks
30mOperationalizing Machine Learning
34mAugmenting Human Decision Making with Data Science
20mShip your Machine Learning Application #bbuzz
32mScaling Personalization: DevOps at Stitch Fix
29mSeveral People Are Tuning: Data and Machine Learning at
40mEnabling simple ML ops at Stitch Fix
42mAugmenting Human Decision Making with Data Science
19mData science for real problems