ML System Design curriculum

Start here — the syntax and mental models everything else builds on.

1. 1

   Framework & problem framing

   ~16 min · 3 lessons

   Structure any open-ended ML design question in 45 minutes.

   requirementsframingmetrics

   • The ML system design frameworkup next6m
   • Framing business goals as ML tasks5m
   • Offline vs online metrics5m
2. 2

   Data & feature pipelines

   ~16 min · 3 lessons

   Design reliable labels, features and data pipelines.

   labelsstreamingfeature store

   • Training data, labels & sampling6m
   • Batch vs streaming features5m
   • Feature stores & training-serving skew5m

Combine data from several places and reshape it into answers.

1. 1

   Retrieval & ranking systems

   ~18 min · 3 lessons

   Build feeds, search and recommendations that scale.

   funneltwo-towerANNre-ranking

   • The two-stage recommendation funnel6m
   • Embeddings, two-tower models & vector search6m
   • Ranking, re-ranking & business logic6m

The techniques that show up in real interviews and real pipelines.

1. 1

   Serving & latency

   ~16 min · 3 lessons

   Serve models fast and roll them out safely.

   batch vs onlinequantizationcanary

   • Online vs batch inference5m
   • Making models fast6m
   • Safe deployment & rollouts5m
2. 2

   Monitoring & feedback loops

   ~17 min · 3 lessons

   Keep models healthy after launch.

   driftretrainingcold start

   • Drift & monitoring6m
   • Retraining strategy5m
   • Feedback loops & cold start6m