Nare Sakhoyan

Engineer working across the LLM stack — from applied product features to model alignment data. Five years shipping production web systems, two years working directly with language models: LLM-driven evaluation and recommendation products, RLHF/SFT/DPO training data and systematic failure-mode analysis, and agent and retrieval systems built on RAG and vector search. TypeScript-first, Python-capable. Comfortable owning a feature end to end, from data pipeline to product surface.