ML System Design, From Problem to Deployed Model
A model is one part of a system. The design decides what the model predicts, who acts on it, what it costs, and what happens when it fails.
110 minutesBuildPythonMembers
Learning Objectives
- Frame a business problem as an ML task with a prediction, a decision that uses it, and a reason why rules are not enough
- Separate offline, online, business, and guardrail metrics, and compare every model against trivial and heuristic baselines
- Estimate the daily cost, latency, and staleness of batch and online inference from traffic numbers
- Choose a rollout pattern (shadow, canary, A/B test) and a safe default for each failure mode
- Identify feedback loops in which the model changes the data that trains its next version
- Write an ML design document and score it against a fixed checklist