← All challenges
Level · 10 challengesLive

Intermediate

Ten challenges — 5 ML, 5 AI — on the habits that separate a demo from a real model.

  1. Challenge 1Scale to avoid leakageIntermediateFit the scaler on training data only, then transform both splits.~20 min · 1 blanks→
  2. Challenge 2Cross-validateIntermediateEstimate accuracy honestly with 5-fold cross-validation.~20 min · 1 blanks→
  3. Challenge 3Precision, not just accuracyIntermediateMeasure macro-averaged precision across all classes.~20 min · 1 blanks→
  4. Challenge 4Build a pipelineIntermediateChain scaling and the classifier into one Pipeline.~20 min · 1 blanks→
  5. Challenge 5Which feature matters most?IntermediateRead feature importances from a random forest.~20 min · 1 blanks→
  6. Challenge 6Teach by exampleIntermediateFew-shot prompting: label examples so the model copies the pattern.~20 min · 2 blanks→
  7. Challenge 7Extract to JSONIntermediatePull structured fields out of a sentence as JSON.~20 min · 1 blanks→
  8. Challenge 8Semantic searchIntermediateRank documents by embedding similarity to a query.~20 min · 1 blanks→
  9. Challenge 9Get a JSON listIntermediateAsk the model for a list inside JSON and parse it.~20 min · 1 blanks→
  10. Challenge 10Constrain the formatIntermediateForce a clean comma-separated list you can parse.~20 min · 1 blanks→