Team Teardown: Organizing and Supporting ML
summary
Airbnb runs an internal ML platform called Bighead (launched October prior year) that teams use for notebooking, feature engineering, model exploration and serving; adoption is partial but migrating legacy models is underway. Top asks are a production feature store, built-in hyperparameter tuning (Zipline) and model observability/interpretability. Some teams build interim, domain-specific tooling and either co-develop with the platform or plan promotion back into Bighead via shared event logging (Jitney).
AI-written summary · about
chapters (18)
- 0:00Introduction and welcome
- 2:19Team teardown panel setup
- 3:43Panelists' roles at Airbnb
- 6:50ML platform history at Airbnb
- 8:37Marketplace intelligence use cases
- 10:22Trust team’s ML challenges
- 11:47Platform user maturity spectrum
- 14:17Platform dependency and features
- 16:45Trust’s platform services
- 18:10Development vs platform adoption
- 20:15Platform as an internal product
- 22:02Fluid relationship with platform
- 23:49Legacy model migration
- 24:52Top platform priorities
- 26:37Prioritizing platform features
- 29:06Feature ownership conflicts
- 30:51Desired platform improvements
- 32:17Closing remarks
related talks
17mLaying the foundation for deep learning
29mEvolving the ML Platform Organisation at Netflix: A Case Study
33mExecutive Summit Roundtable Discussion
30mTeam Teardown: Driving Platform Adoption and Success
33mBuilding Resilient (ML) Pipelines for MLOps
22mBefore the Model: How Machine Learning Products Start...
26mManaging ML Models @ Scale
28mBuilding Developer-Centric ML Inference Platforms