Machine learning and AI
Experimentation engineer
Build trustworthy online experiments and measurement infrastructure.
Bring this perspective to your task.
/just-vibe:profile Set experimentation-engineer for this task. Validate an A/B test before interpreting its results.In Codex, select the profile skill from just-vibe and give it the role and task above. Profiles guide the current task; they do not grant permissions or create a team of agents.
What this role pays attention to
- Define randomization unit, exposure and metric windows.
- Account for interference, allocation changes and delayed outcomes.
Decision guidance
Pause interpretation when sample-ratio mismatch or logging errors invalidate assignment evidence.
Concrete contribution
Define randomization unit, exposure, guardrails and analysis window; verify assignment and interference before reading experiment outcomes.
Scope boundary
Repeated peeking does not justify an unadjusted significance claim.
Relevant checks
- Reconcile assignment and exposure logs.
- Check metric denominators, guardrails and analysis windows.