ML/AI engineer and CS undergrad at NYU Tandon (B.S. Computer Science, May 2027). I build multimodal retrieval, VLM pipelines, and agentic systems, and I ship them end to end.
LinkedIn · New York, NY
- Machine Learning Engineer, Cato Digital (Oct 2026 - present): bare-metal NVIDIA GPU bring-up for datacenter deployment.
- Undergraduate Researcher, NYU Self-Drive VIP (AI4CE Lab, Prof. Chen Feng): hardening an Earth Rover Challenge navigation stack (SegFormer + monocular depth + pure pursuit).
- Project Manager, NYU Stern BAC: leading 4 engineers on VibeScent, a multimodal fragrance recommender (Qwen VL embeddings + BM25, VLM reranking on NVIDIA NIM).
- Project Manager, Google Developer Group @ NYU: AI Companion, a Chrome extension that turns course material into RAG-grounded answers and spaced-repetition quizzes.
Previously: Member of Technical Staff at VaultIQ (video-to-text VLM pipeline), Software Engineer Intern at SkyNet Worldwide Express.
| Project | What it is |
|---|---|
| Thermal Commons · live app | 8,760-hour Python heat-recovery model + Next.js/MapLibre siting app for data-center waste heat. 3rd place, BAC x iMasons Hackathon. |
| Agora | Voice-first Socratic tutor: LangGraph state machine, Whisper STT, Qdrant RAG. Best First Time Hack + Best Use of Vultr at HackNYU. |
| Hinglish NER | Fine-tuned XLM-R / mBERT on 18k+ code-mixed examples vs. zero-shot LLMs (LLaMA 3.1 8B, GPT-4o). |
| nexus.ai | Multi-agent bull / bear / judge debate for investment research over a Monte Carlo engine and DuckDB. |
| Glass-Box Connect 4 | NumPy MLP policy + bitboard alpha-beta solver, graded against perfect-play labels. |
| Rainbow DQN from scratch | Rainbow-style Deep Q-Network on CartPole in pure NumPy, no PyTorch. |
Merged PRs into Astro, h3, Solid docs and cavemem.
ML: PyTorch, NumPy, scikit-learn, LangChain / LangGraph, RAG, fine-tuning, VLMs, FAISS, Qdrant, pgvector Systems: Python, TypeScript, C++, FastAPI, PostgreSQL, Redis, DuckDB, Docker, AWS, CI/CD Web: React, Next.js, Node, WebSockets