At Turing, I evaluated and annotated LLM outputs across complex reasoning and coding tasks, comparing model responses and documenting the technical rationale behind my judgments. That work contributed RLHF data aimed at improving reasoning, calibration, and alignment.
At Engagement Agents, I review scrape output against live sources, tracing failures to their root causes and documenting them field by field with evidence. I reviewed 280+ jobs in one week and proposed a “Not assessed” status so unchecked fields wouldn’t skew accuracy data.
I built AIWA Codes, an agentic platform that turns prompts into full-stack apps and runs automated checks on each build. Its browser agent tests real user flows, while the evaluation pipeline checks plans, API coverage, and task contracts.
