MLOps for High-Stakes Environments
summary
Toyota Research Institute builds an MLOps pipeline that turns vehicle logs into labeled scenarios, trains models, runs quantized TensorRT builds on hardware-emulating racks, and gates releases with simulation, regression suites and closed-loop safety certification before limited fleet staging. The stack enforces one system two modes (chauffeur vs guardian), tracks model and system metrics including kilometer per disengagement, uses open and closed loop checks plus humans-in-the-loop, and prioritizes scenario-driven iteration to scale safe fleet learning.
AI-written summary · about
chapters (14)
- 0:00Introduction and welcome
- 0:30About Toyota Research Institute
- 1:12One system two modes approach
- 2:37Machine learning in automated driving
- 4:25Building generalizable ML models
- 6:54MLOps for automated driving
- 9:02Challenges in MLOps
- 10:45Model versus system performance
- 13:15Complex model dependencies
- 15:01Safety in MLOps
- 17:09Safety certification process
- 17:31Accelerated model iteration
- 19:40Future of ML at TRI
- 21:06Q&A session
related talks
13mMLOps at Scale
30mMLOps for Continuous Learning
13mHuman-in-the-Loop MLOps: Boost Model Performance by 40%
53mExperience with an end-to-end MLOps pipeline: Practical insights and 3D architectural perspectives
43mNavigating the MLOps Journey: Key Considerations for Successful Implementations
13mMLOps: From AI Research to Production-Ready Systems
47mMachine Learning in Production: An Intro to MLOps.
25mBuilding Scalable MLOps Pipelines That Actually Work