AI Observability: How To Fix Issues With Your ML Model
summary
Argues that detecting drift is not enough and that logging is the key to root cause analysis, using WhyLogs profiles that compress terabytes to megabytes, record distributions and nulls, and merge across time or nodes to preserve historical state. Shows a working pipeline where an AI observability platform ingests logs, raises alerts, and an automated MLOps actuator either retrains models on new data or rolls back a broken data pipeline, turning slow human remediation into sub-minute automated fixes.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:36What Is a Machine Learning System
- 1:48Agenda
- 3:10Data Quality Issues
- 7:40The Secret to Observability
- 13:11Human in the Loop Approach
- 13:30Human in the Loop
related talks
10mML-based Log Analysis for Faster Debugging
12mObserving AI: Why Even LLMs Sometimes Need Therapy
01h 00mWhy observability requires the marriage of AI, metrics, and logs
27mSave your ML model from degradation with data logging
13mHuman-in-the-Loop MLOps: Boost Model Performance by 40%
01h 21mBuilding a ML Model Observability pipeline
28mWhere are we going with observability?
42mMachine Learning In Observability Systems: From Insights To Action