← All challenges
Track · 7 daysLive

7 Days of MLOps

Package, deploy, and monitor a model end to end, one step a day.

  1. Day 1Package a trained modelBeginnerTurn a trained model into a portable artifact you can save, ship, and load anywhere.~25 min · 2 blanks→
  2. Day 2Serve it behind an APIBeginnerWrap the model in an inference function: raw input in, a clean JSON-ready response out.~25 min · 2 blanks→
  3. Day 3Validate the inputIntermediateGuard the boundary: reject malformed input with a clear error before it ever reaches the model.~30 min · 2 blanks→
  4. Day 4Tests that gate the modelIntermediateCI for ML: a behavioural test plus a performance gate that block a bad model from shipping.~30 min · 2 blanks→
  5. Day 5Monitor in productionIntermediateLog prediction confidence and watch the live distribution — your early-warning system.~30 min · 2 blanks→
  6. Day 6Detect data driftAdvancedCatch the silent killer: incoming data drifting away from what the model trained on.~30 min · 2 blanks→
  7. Day 7Automate retrainingAdvancedCapstone: close the loop — detect a drop below SLA, retrain on fresh data, verify recovery.~35 min · 2 blanks→