From experiment to production - a journey of a machine learning model
summary
Shows a concrete ML pipeline that takes an IMDB sentiment model from experiments through CI/CD to production, using Weights & Biases for experiment and artifact tracking and GitLab CI + Airflow for release and orchestrated training. Models retrain on CI to ensure up‑to‑date data, artifacts and evaluation metrics register with versioned tags, deployment packages model plus function code, and production monitoring logs input distribution (e.g., average review length), prediction confidence and accuracy and auto‑triggers retraining on degradation.
AI-written summary · about
chapters (8)
- 0:00Introduction to ML journey
- 1:47What is MLOps
- 3:33Experiment tracking in ML
- 8:33Model evaluation process
- 10:23Model release pipeline
- 14:38Model deployment workflow
- 16:44Model monitoring in production
- 20:54MLOps takeaways
related talks
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
35mHow to deploy machine learning models to production (frequently and safely)
57mDevOps for Machine Learning
57mManaging Machine Learning Lifecycle with MLflow
38mFrom Experimentation to Products: The Production Machine Learning Journey
01mintro video
28mLeading machine learning research journeys with happy (into-production) endings
25mContinuous ML Applications in Production