Working where machine learning meets the hardware it runs on.
I'm a CS student at the University of Massachusetts Amherst (Class of 2028, 4.0 GPA). My focus sits at the hardware–software boundary, GPU computing, ML infrastructure, and performance, where understanding the workload and the machine underneath it both matter.
I split my time across two roles:
- Undergraduate Researcher, ML4Ed Lab — studying how LLM pipelines fail under long-context conditions, and building benchmarks to measure it.
- AI/Software Engineering Intern, Commonwealth of Massachusetts (EOTSS) — building a VLM-based image tagging and natural-language search system over ~90K government images on AWS; migrated the stack to Lambda/API Gateway via CDK and handled the latency and infra work that came with it.
The throughline: I like going one layer down.
🏆 Wins: HackUMass XIII · Hack(H)er413
🔭 Now: building a scalable multimodal AI search platform on AWS, enabling distributed VLM image tagging and text-to-SQL search across ~90K government images (EOTSS)
🌱 Learning: low-level C++ performance · the systems behind fast ML · interpreters · GPU architecture & optimization (CUDA)
🔐 DataVault — Winner, HackUMass XIII A privacy-preserving RAG pipeline for natural-language querying over encrypted data, with zero-memory database handling so sensitive records are never persisted in plaintext. (Python · Docker · PostgreSQL)
📈 Project Oracle A cloud-hosted multi-agent system for quantitative trading, delegating between Manager, Researcher, and Analyst agents over a vector store. (AWS · C++ · Pandas)
🛡️ Project Sentinel A cybersecurity agent using an XGBoost model (98.7% recall) to detect network threats in real time and generate human-readable reports via the OpenAI API. (React · Flask · scikit-learn)
Systems & Languages
Infrastructure & Data
Web
