Learn/Course 19
Production · weeks 23

Reliability, monitoring & experimentation

Model quality decays while every infrastructure dashboard stays green.

Opening production brief

A measurable problem, before a model.

Users

Operators, product owners, and people affected by the reliability decision.

Data

SLIs/SLOs, delayed labels, drift, replay tests with version, owner, event time, and availability contracts.

Constraints

Model quality decays while every infrastructure dashboard stays green. Every solution must state latency, cost, capacity, and fallback limits.

Success

Offline evidence plus a deployable resolution: Joined data/model/service telemetry, drift policy, experiments, alerts, and runbooks.

CHAPTER 1

Frame and baseline

Turn the production problem into explicit data, metric, and baseline contracts.

  1. 01

    ML SLIs, SLOs, error budgets, ownership

    ML reliability joins immediate system health with delayed decision truth under one owned operating contract.

    65–90 min · checkpoint · deterministic lab
  2. 02

    Observe system, data, prediction, and business

    Observability becomes diagnosis only when every layer can be joined by exact version, request, time, and slice.

    65–90 min · checkpoint · deterministic lab
CHAPTER 2

Build and stress

Formalize the machinery, test counterexamples, and expose system limits.

  1. 03

    Diagnose, mitigate, and write postmortems

    Mitigate harm quickly, preserve causality carefully, and make the postmortem change the system rather than the story.

    65–90 min · checkpoint · deterministic lab
  2. 04

    Randomized experiments and guardrails

    Trust an experiment only after assignment, exposure, unit of analysis, uncertainty, and guardrails all agree.

    65–90 min · checkpoint · deterministic lab
CHAPTER 3

Resolve and operate

Join model behavior to architecture, observability, rollout, and rollback.

  1. 05

    Drift, retraining, rollback, continuous verification

    Continuous verification should automate evidence collection; promotion remains a gated response to demonstrated decision need.

    65–90 min · checkpoint · deterministic lab
  2. HW

    Launch analysis and incident response

    Analyze a synthetic launch, verify assignment, estimate effects, inspect slices, and choose continue, pause, or rollback.

    Rubric · staged hints · reference resolution
Prerequisite guidance

3 concepts make this faster.

Prerequisites are guidance, never hard gates.

C02EvaluationA metric contract, leakage-safe validation, calibration analysis, and experiment plan.C17ML platformsA traced batch/stream platform with orchestration, registry, lineage, and ownership.C18Serving systemsBatch and online paths with batching, autoscaling, load tests, canaries, and rollback.
Ready to work the problem?

ML SLIs, SLOs, error budgets, ownership

Start lesson 01