DeepBrew – AI
summary
A production platform called DeepBrew runs on Azure with a datalake, Databricks ETL, Docker/AKS, Git + Sonar + CI/CD and preapproved PII access to avoid repeated security reviews. It serves billion recommendations per year to 40 million active customers, is live in 3,200 stores for a digital drive‑through, and returns results in sub-200 ms. Models use reinforcement learning DQN and Thompson sampling; outcomes focus on shorter drive‑through queues, safety, and operational reliability.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:16Seven Deadly Sins
- 0:45Starbucks Promise
- 1:14DeepBrew
- 2:29Dynamic Yield
- 3:23Deep Brew
- 5:04Deadly sins
- 6:06Making claims
- 9:00Leadership support
- 10:17Security
- 11:22Operations
- 12:57Agile
- 14:16Innovation
- 17:52QA
- 18:10Agile vs Waterfall
- 22:17Software Process for New Features
- 28:52AI Center of Excellence
- 30:02Platform
- 36:36Azure ML
- 42:11Spotify Partnership
related talks
40mBuilding secure and efficient AI solutions in Azure with Partners
01h 16mAI Advantage Reinventing SME Operations
37mHarnessing OpenAI to help millions
40mInsights and Key Learnings on Our AI Journey
35mDeepL: A Product Culture that Creates Top-Notch AI Products
50mAzure AI, Power new possibilities for every organization
46mPower AI apps and develop rich experiences with Azure SQL Database
37mOn Building Production-Quality AI Systems