At Ethara AI, I evaluate robot manipulation clips and score atomic actions, movement, and overall behavior so models receive feedback at multiple levels. I’ve evaluated 350+ clips and prepared frame-by-frame descriptions, maintaining an average quality rating of 4–5 out of 5.
As an LLM post-training Intern at Ethara AI, I evaluated 40–50 tasks per week against quality criteria and assessed frontier LLMs against standard benchmarks. I also tested how prompt wording and structure affected model outputs, documenting patterns that produced accurate, consistent responses.
I built a multi-turn RAG agent for the Conversational Assessment Recommender that asks clarifying questions before recommending and comparing assessments from a 369-item catalog. I also built an inventory and order management app with FastAPI, Docker, Railway, and Vercel to automate stock updates and order-total calculations.
