AI Trainer — Code & Reasoning Evaluator | Revelo (Contract)
Served as an AI Trainer and Code & Reasoning Evaluator, assessing AI-generated solutions paired with prompts. Evaluated the quality, correctness, and completeness of both the reasoning trace and the final code output for engineering-grade CodeGen tasks. Applied turns taxonomy to multi-turn interactions by categorizing each conversation turn to identify reasoning gaps, off-track responses, and quality degradation patterns across depth.• Labeled and categorized multi-turn conversation turns to flag reasoning and quality issues.• Reviewed AI-generated diffs for logical correctness, code quality, and test coverage adherence.• Wrote evaluation rubrics and annotation guidelines to standardize scoring across task types.• Provided structured feedback to reinforce desired model behavior on realistic engineering prompts.