MLOps for Continuous Learning
summary
Argues that MLOps is a cultural practice plus a toolchain and lists concrete components: PyTorch, TensorFlow, Ray for training; ONNX + ORT to standardize and accelerate cross‑framework deployment (works with Azure ML); MLflow for tracking, projects, models and a registry; Kubeflow/Airflow/Jenkins for pipeline orchestration. Recommends pipelines include data prep, parameterized training, repeatable validation, versioned deployment, and monitoring/alerting so models can be continuously retrained and debugged.
AI-written summary · about
chapters (10)
- 0:00MLOps and continuous learning
- 3:17Why MLOps matters
- 5:49Core machine learning operations capabilities
- 7:13Open source MLOps tools
- 11:32Model deployment
- 15:27ML pipelines for learning
- 18:34MLflow platform
- 21:04Resources and closing
- 21:45Q&A on ONNX and AI careers
- 25:41Building MLOps team
related talks
45mThe Culture, Architecture and Technology of MLOps
14mWhy Your DevOps Skills Are Your MLOps
13mMLOps at Scale
13mMLOps: From AI Research to Production-Ready Systems
20mBringing CI/CD Practices to Machine Learning with MLOps
36mIs the MLOps tooling the same as the DevOps tooling?
01mMLOps: Applying DevOps to ML
41mTHE STATE OF MLOPS - MACHINE LEARNING IN PRODUCTION AT ENTERPRISE SCALE