Measure workload health, behavior, evaluation quality, capacity, cost, incidents, and operational evidence.
Learning outcomes
Curriculum
The chapters are individually addressable documentation pages. You can link directly to a concept from another program, architecture decision, or implementation guide.
Define operational outcomes and the evidence required to explain them.
Operate AI as a measurable service whose technical health, behavior, risk, and cost can change independently.
Design signals that explain service health, AI behavior, policy outcomes, change, and incidents without leaking sensitive content.
Design evidence alongside controls so important decisions, releases, configurations, and operating events can be verified.
Measure service health and AI behavior across the request lifecycle.
Set service objectives and prepare for degradation, incidents, and recovery.
Define meaningful objectives and operational responses for degraded, unsafe, or unavailable AI behavior.
Define capacity envelopes, degradation behavior, service objectives, recovery targets, and tested restoration evidence.
Connect demand, capacity, performance, cost, and accepted outcomes.
Measure constrained capacity, demand, waste, unit cost, and accepted outcomes without optimizing spend in isolation.
Version complete AI behavior, assess change impact, approve promotion, support rollback, and retire superseded assets.
Record context, options, control consequences, evidence, dependencies, limitations, and reversal triggers for material decisions.
Create an operations package and verify the module outcomes.
Create service objectives, incident evidence, capacity measures, and unit economics.
Design exercise; executable implementation pending
Run a synthetic local retrieval system that enforces document permissions, citations, deletion, no-answer behavior, and injection controls.
Runnable synthetic prototype with recorded validation
This curriculum supplies an operating method, not universal SLO targets, cost forecasts, or a production monitoring configuration. Workload baselines, thresholds, retention, and tool-specific dashboards require measured local validation.