Team Teardown: Evolving an ML Platform
summary
Centralized ML team of about 7–8 engineers and 5 data scientists builds a serving platform, model management and a forthcoming feature store; onboarding moves from whiteboard handoffs to a Slack/API flow and templates cover ~95% of models, cutting mean time to deployment by ~11x and halving cost to add a model. Platform adds auto‑monitoring, shadow testing and auto‑retraining, runs in a hybrid cloud, and requires tight collaboration with data engineering to standardize feature definitions; remaining gaps are education and saying no to unsuitable requests.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:30Meet the Team
- 3:33Role of Platform
- 12:08Onboarding Process
- 15:57Customer Perspective
- 16:58Managing Expectations
- 18:13Ideal Execution Map
- 19:09What is still hard
- 20:34Education
- 21:52Challenges
- 22:44How to make a small team cohesive
- 23:43How to make a complete platform
- 24:42Roadmap
- 27:25Data Engineering
- 27:53Data Engineering Collaboration
- 28:58Central Feature Store
- 29:39Metrics
related talks
30mTeam Teardown: Driving Platform Adoption and Success
29mEvolving the ML Platform Organisation at Netflix: A Case Study
29mBuilding an ML Platform from Scratch
09mPlatform Engineering for Data Science at Scale
28mBuild an End-to-End ML Workflow for Your Organization
28mBuilding Developer-Centric ML Inference Platforms
26mManaging ML Models @ Scale
28mPractical steps to establish and scale your data & ML team