At Ethara.AI, I evaluated and improved LLM responses through RLHF, prompt evaluation, response ranking, and annotation to support model alignment. I assessed outputs for accuracy, reasoning, coherence, instruction-following, and safety.
On CampusConnectAI, I built a RAG pipeline with FAISS and MiniLM-L6-v2 embeddings, improving semantic retrieval accuracy by 65% over keyword-based search. I also developed Aura-CX, a multi-agent commerce platform, and an OpenCV-based face recognition attendance system.
