
A bit of everything on the systems side — pipelines, database migrations, network stuff. Whatever infra problem shows up, I'm probably the one on it.

Built a neural net that acts like a fast digital twin for a much slower path-dependent volatility model, getting close to the accuracy of the full model while making calibration fast enough for real-time use. Advised by John Dodson.

Built the company website pretty much solo, working directly with the CEO. Nice bump in engagement and retention after launch.

Freelanced full-stack projects for a handful of clients. Good way to learn fast and get paid doing it.
Quant research, trading systems, and a few things built just to build them:
1st place — FMA Quantitative Trading Competition
An event-driven trading bot I built and live-traded for the competition — momentum signals, automatic stop-losses, the works. Judged by folks from DRW, Cargill, and OCC, and it never lost money live.
Undergraduate derivatives research
A neural net that mimics a much slower options-pricing model — basically trying to get the speed of a simple model with the accuracy of a complicated one. Added guardrails so it knows when to stop trusting itself and fall back to the full model.
OOP backtesting engine
A little framework for testing trading strategies before risking real money on them. Has an optional news-sentiment add-on.
Self-hosted always-on daemon
Scrapes job postings from a bunch of sources, filters out the noise, and pings me when something good shows up. Runs on its own 24/7.
Sim-racing hardware
Built a handbrake for sim racing from scratch — potentiometer, Arduino, and a hand-welded steel lever. Plug-and-play with any racing game.
Degree-planning tool for UMN students
Built this with some friends from the Social Coding club. Pulls real course data so you're not planning off stale info.
Konrad Trestka