Arc 03 · AI Engineer
Learning Systems and Classical ML
Build the discipline that still governs foundation-model products: splits, baselines, leakage control, calibration, drift, and causal humility.
Exit capability: Ship and defend a supervised learning system with trustworthy evaluation.
- Mapped lessons
- 7
- Published now
- 0
- Full-arc estimate
- ≈15 hours
- Last edited
- 2026-08-11
Lesson sequence
Live units open into complete labs. Planned units stay visible to show the dependency path, but intentionally have no detail route.
- 01
The Generalization Contract
PlannedDefine the population, task, loss, and deployment distribution a model is expected to generalize to.
Concept labFoundation80 min estimate
- 02
Split Data by Causality, Not Convenience
PlannedUse time, identity, groups, and provenance to prevent leakage that random splits conceal.
Failure labIntermediate95 min estimate
- 03
Linear Models as Debugging Instruments
PlannedUse linear and logistic regression to expose data and target problems before adding model complexity.
Build labIntermediate100 min estimate
- 04
Trees, Boosting, and the Shape of Residual Error
PlannedUnderstand how tree ensembles partition mistakes and when their inductive bias wins.
Build labIntermediate110 min estimate
- 05
Calibration, Thresholds, and Decision Cost
PlannedTurn scores into actions using reliability curves and asymmetric error costs.
Systems labIntermediate100 min estimate
- 06
When the Product Needs Ranking, Not Classification
PlannedChoose ranking losses and metrics when ordering quality matters more than labels.
Concept labIntermediate90 min estimate
- 07
Drift, Feedback Loops, and Delayed Labels
PlannedDesign monitoring when the system changes the data it later learns from.
Systems labAdvanced110 min estimate