Tutorials

Four case studies, one shared arc — the case → the data → problem formulation → features → run → evaluate — modeled on DSSG triage’s Dirty Duck tutorial and extended with the surfaces triage-pg adds (dashboard, in-PG fairness and audition, subset evaluations, monitoring, survival).

You are…	Start with	You’ll learn
checking your setup works	Dirty Duckling	a pass/fail smoke test + 5-minute dashboard tour
learning temporal ML on policy data	DirtyDuck	the full case study — early warning vs resource prioritization on the same data
running something production-shaped	Chicago 311	fairness auditing, subset evaluations, monitoring, survival
deep into feature engineering	DonorsChoose	Deep Feature Synthesis over a multi-stream entity graph

Read them in order for the full course — smoke test, the two observation regimes on one dataset, the production surfaces, the three signal regimes — or jump to the row that matches you. The onboarding one-pager is the system-at-a-glance map, and the runbook READMEs in the repository stay the terse command references.