CourseML System Design · 0120 minPremium
Framing and metrics
Turning a product goal into a learning problem: choosing the prediction, the objective and the decision built on it, then the three layers of metrics (business, online, offline), what each offline metric actually measures, calibration, thresholds from costs and the proxy metrics that quietly reward the wrong behaviour.
Premium
Part of the ML System Design course, on the Premium plan
Readers with Lifetime Access read it in full. Already bought it? Sign in from the top bar.
What’s inside
- 1. From product goal to ML task
- 2. Three layers of metrics
- 3. Offline metrics, and when each one lies
- 4. Calibration
- 5. Thresholds come from costs
- 6. Proxy-metric traps
- 7. Requirements that change an ML design
- What goes wrong
- Drills
- Check yourself