About Me
I studied computer science and physics at Northeastern because I was drawn to both how the world works and how we can build systems to model it. I graduated magna cum laude, but more importantly the combination shaped how I approach problems: start from first principles, understand what is actually happening, and build something practical.
Wayfair was where I first got a real feel for production engineering at scale. As a software engineer co-op, I worked on Python FastAPI endpoints for internal platform tools and automated CI/CD across enterprise repositories. It was a good introduction to the difference between code that works in isolation and code that other teams rely on every day.
Around that time, I also spent six months at Forschungszentrum Jülich PGI-14 in Aachen as a research intern. I developed C++ and Python libraries for neuromorphic hardware solvers and benchmarked optimization algorithms on a 256-CPU HPC cluster. That work led to a co-authored IEEE ICRC 2024 paper.
From there I joined Pacman HEF, a construction technology startup, where I worked on an AI-powered takeoff tool for electrical estimators. I built a computer vision system to detect and count electrical symbols across full sets of construction plans, with the potential to cut manual takeoff time by up to 5x. I also helped build the full-stack platform around it with React, Express.js, PostgreSQL, Docker, and Kubernetes.
I'm currently an AI Engineer at Hyper Company, where I work on production multi-agent systems for enterprise clients. A lot of the job is closing the gap between something that looks impressive in a demo and something that is reliable enough to run as part of a real business process.
In practice, that means owning end-to-end workflow automations across document processing, decision routing, and exception handling at meaningful scale. For one insurance client, I built an automation that processed nearly 1,900 new business submissions with zero human review, and I've worked on other pipelines that handle tens of thousands of executions per month with 80-99% straight-through rates. Most of the work is careful iteration: finding where systems break, tightening the edges, and making them dependable.
I'm most interested in problems where software has to operate in the real world, with messy inputs, edge cases, and real consequences when things go wrong. Agentic systems are a big part of that right now, and I think we're still early in learning how to build them well.
Outside of work, I still follow quantum computing, optimization theory, and computational neuroscience. Those interests are part of the same thread for me: understanding complex systems well enough to make them useful.
You can find more of my work on GitHub, or reach out on LinkedIn.