Scaling ML platforms to industrialize AI/ML development
summary
Company verifies millions of identities daily using hundreds of ML models and builds a proprietary internal data labeling tool that becomes impossible to maintain; market research finds over 100 data labeling tools in 2022; discovery process uses user interviews, architecture mapping and multiple POCs; hard constraints (PII, on‑prem storage, vendor geography) and stakeholder politics force reprioritization; decision timeline stretches over three months and ultimately narrows to two tools that fit constraints.
AI-written summary · about
chapters (5)
- 0:00Introduction
- 3:40ML platforms at scale
- 6:54Beyond technical needs
- 10:26Process design & alignment
- 13:18Summary
related talks
34mScaling ML in the Traditional Enterprise
46mPanel: OpML
04mManaging Machines
32mTeam Teardown: Evolving an ML Platform
40mData labelling to scale your ML and AI pipelines
30mChallenges of machine learning: The new ML lifecycle
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
31mUse-Case Driven ML Platforms