AI engineer building agent tooling — shipped the Claude Code plugin integration and exact-person recovery at HiRey
I'm doing a Masters in AI at Northeastern and interning at HiRey, in Boston. Most of what I build is retrieval and agent systems. **RecruitIQ** — a RAG recruiting assistant. FAISS vector search over a resume and job-description corpus, Groq `llama-3.1-8b-instant` for generation, `all-MiniLM-L6-v2` embeddings, orchestrated in LangChain behind a Streamlit UI. Semantic candidate matching, free-form recruiter Q&A over the corpus, and structured gap analysis against a specific JD. **Memory Chat** — a Next.js chat app that remembers you across sessions. Every turn runs recall → generate → remember, so context persists instead of resetting. At HiRey I'm working on the People surface — the path from a first clue to an exact person — writing regression tests around it. Happy to talk to anyone building with agents, especially about what breaks in front of real users: grounding, retrieval quality, evaluation.