AI Data Annotator and AI Trainer

Real portfolio pieces built during the course, not a certificate for its own sake. Each one is work you can show.

• Intent label logA twelve-item customer message batch labeled against a versioned guideline, with the rule that decided each label, every required flag, and a reasoning note that never repeats sensitive values.
• Calibration reviewThree calibration misses taken apart by cause before action: one explained by a guideline change and accepted, one gold error disputed with the rule quoted, one private habit named and logged, and the score stated exactly.
• Preference justificationsFour AI response pairs scored on four rubric dimensions against a client policy card, with hard rules applied, a preference chosen on a five-point scale, and the single deciding difference named for each, including a genuine tie.
• Audit response and error logA response to five quality audit findings: two disputed with a changelog time and a label definition, three accepted with the habit behind each and a concrete prevention step, the data handling finding treated first, and a pace request answered without breaking the guideline.