ML Systems
How do we deploy machine learning in production?
Organise data, labels, feature pipelines, serving, monitoring, retraining, and experiments into reliable production systems.
1. Problem Definition
Define the product goal, model goal, and failure modes for YouTube comment toxicity detection.
2. Data Collection
Collect the examples, logs, and context needed for a production model.
3. Labels
Design reliable labels while handling ambiguity, disagreement, and noise.
4. Feature Pipelines
Compute features consistently for training and serving without leakage.
5. Training Pipelines
Turn data, code, configuration, and evaluation into repeatable model training runs.
6. Model Serving
Deliver model predictions with latency, reliability, and fallback constraints.
7. Monitoring
Watch deployed models for drift, latency, reliability, and failure patterns.
8. Retraining
Refresh models safely as data, users, and product behavior change.
9. A/B Testing
Measure production impact before fully launching model changes.