Responsible Engineering: Evaluation, Governance & Auditability
Four weeks turning an existing AI system into one you can defend: evaluation harnesses, approval gates, audit trails and the evidence a regulator or auditor will actually ask for.
Join the waitlistWho should take this program?
You'll get the most from this if
- You work with an AI system already in use or close to launch
- Comfortable reading code, even if you do not write it daily
This is not for
- Teams with nothing built yet — build first, then govern
What you will build
- An evaluation harness with a golden dataset, running in CI
- An approval-gate design for your highest-risk action
- An audit trail that reconstructs a decision end to end
What you learn, week by week
Module 1 — Building the golden set
- Mine real failures instead of inventing examples
- Label consistently
- Version the dataset like code
Module 2 — Assertions and judgement
- Write deterministic assertions where behaviour must be exact
- Use LLM-as-judge where it cannot, and calibrate it
- Run everything in CI
Module 3 — Gates and thresholds
- Place approval gates where a wrong action is irreversible
- Set confidence thresholds that escalate
- Design the human handoff so it is actually used
Module 4 — Audit and evidence
- Log enough to reconstruct any decision
- Produce the evidence pack an auditor asks for
- Write the model card and keep it current
How this program handles evaluation and governance
The entire course is the responsible engineering component. The distinction it insists on: governance produced after a system is built is documentation, and governance designed into the loop is engineering. You leave with running code, not a policy document.
Tools and stack you will use
Fees
Request pricing
Fees depend on cohort, format, and whether this runs for an individual or a team. We'll send the full breakdown.
Get fee detailsQuestions people ask
Is this a compliance course?
No. It is an engineering course that produces the artifacts compliance needs.
Do we need a system already?
Yes, or a realistic one. Governing a hypothetical teaches nothing.
Does this cover the EU AI Act?
We cover what it requires of your engineering. Legal interpretation belongs with your counsel.