I'm wrapping up my MS in Computer Engineering at NYU (May 2026). My work sits at the intersection of agentic AI, LLM fine-tuning, and applied ML — some of it in research labs, a lot of it built fast at hackathons and then pulled apart afterward to figure out why it worked.
Two things I optimize for: models that survive contact with real data, and shipping quickly enough to actually learn something.
- 🥇 Winner, Best Use of Hugging Face — Mistral AI Worldwide Hackathon '26 Agentic World (Nvidia, Mistral AI) — behavioral digital twins that learn user archetypes from behavioral signals (clicks, scroll velocity) built on Mistral Small 3.1 + Nvidia Nemo + Nvidia 49b + Codestral, with per-demographic Mistral-7B LoRA adapters trained on real session data. All models open-sourced on Hugging Face.
- 🥇 3rd Place Overall, Best Use of Blaxel(YC'25) — Iterate × Columbia Hack '26 An MCP adapter sandbox that turns OpenAPI specs into serverless tools, automatically.
- 🥇 Winner, Transit Track(NJ Transit) — Rutgers Hackathon NJ Transit AI — Ensemble delay prediction on live transit feeds (87% accuracy) over an end-to-end streaming pipeline serving sub-second alerts.
- 🥈 Runner-Up, Best Use of Daedalus Labs(YC'25) — Columbia DevFest '26 A seven-stage pipeline that mines, synthesizes, and deploys production-ready MCP servers.
- Research Assistant — NYU VIDA Lab (Visualization & Data Analysis) — captured synchronized audio-video from pedestrian sensor arrays and automated the collection pipeline with SSH-driven Python scripts.
- Research Assistant — NYU WIRELESS — tested 7–24 GHz signal propagation using LiDAR sensors and adjustable antennas, characterizing how mid- and low-range signals pass through different media and how to capture them most cleanly.
- Co-author, CVPR 2026 — PedTiles-6K: A Dataset for High-Resolution Pedestrian-Infrastructure Segmentation
Grapher Labs — co-founder of a B2B fintech startup building a causal financial simulation engine for CFOs. Evaluated product market fit and signed LOIs with early customers. Selected for and completed the NYU Leslie eLab accelerator.
Languages Python · C++ · Java · JavaScript · R ML / AI PyTorch · TensorFlow · Hugging Face · LoRA fine-tuning · scikit-learn · NLP · Computer Vision Data / Streaming Kafka · PostgreSQL · BigQuery · Dataflow · Pub/Sub · ETL pipelines Platform Docker · Kubernetes · AWS · FastAPI · React · Node.js