AI Trainer and Reviewer | Mercor (Contract)
Designed AI training scenarios to coach agents in completing real-world tasks across common communication and productivity platforms. Created prompts and world-state data used to evaluate agent decision-making and task execution against defined success criteria. Reviewed and refined LLM outputs to ensure accuracy, quality, and alignment with the rubric-based goals. • Built personas, artifacts, and rubrics to generate structured multi-layer training datapoints. • Managed task tracking and documentation using Airtable and the Crucible platform. • Ran iterative agent evaluations both unaided and with guided hints to assess performance under varied conditions.