About me

Research, trading, and portfolio management

My career has taken me through the full process of systematic investing: building research tools, developing signals, constructing portfolios, and trading them live.

Bridgewater: research tools and investment engineering

At Bridgewater, I worked directly on Backtesting and Simulation as a product, developing its libraries and APIs and addressing bugs and feature requests. As an Investment Engineer, I paired with Investment Associates to implement investment ideas, simulate and analyze signals and estimates, and work with walk-forward models. That brought software engineering and investment research together in my daily work.

Bot Lab: connecting signals to live trading

I then worked on US equities statistical arbitrage at Bot Lab, an external Schonfeld pod, under a lead portfolio manager who had previously been a quant at Citadel. My responsibilities included live order submission and review, transaction cost analysis, portfolio and results analysis, and risk analysis using Barra and custom factors. I worked with underlying and master signals and live portfolio construction, and developed signals from industry datasets under the lead PM’s guidance.

PharVision: responsibility for the whole strategy

At PharVision, I became a portfolio manager responsible for a US equities statistical arbitrage strategy from research through live trading. I designed and implemented the signal research pipeline, working with roughly eight alpha models, typically around 100 engineered features per model and on the order of 1,000 engineered features across the research.

The models used traditional machine learning to forecast residual returns over one or more forward windows. Their themes included IBES and analyst data, news and sentiment, options trades and quotes (TAQ), and fundamental data, alongside shared risk, market, and residual-return features. I also designed and implemented portfolio construction for research and live use, formulating it as disciplined convex programming with CVXPY. This built on my earlier experience with MOSEK at Bot Lab.

AI for quantitative research

Since late 2025, I’ve been exploring how Claude Code and Codex can support repeatable research workflows through reusable skills, retrieval-augmented generation (RAG), and careful context management. I’m interested in AI agents as research collaborators, with human judgment and theoretical reasoning guiding hypotheses and feature design.

The research stages I want to connect include exploring linear relationships, understanding risk, constructing features, scheduling model tuning and refitting, producing forecasts, and evaluating them through disciplined sampling and statistical analysis. Portfolio-construction research is another focus. The appeal is making these stages easier to reproduce, inspect, and build on.

Building beyond finance

I use Claude Code, Codex, and local models through Ollama on my own GPUs in my personal workflows. They make building both more productive and more enjoyable. I’ve been developing a structured platform for my personal projects, with a management dashboard and AI development workflows.

This site is a new project, launched on October 1, 2026. It brings together the books, tools, and projects around my room, including ClaudeD, my music app. I’m planning demos of these workflows as they develop. You can read the project stories and explore the room in the meantime.

Away from the desk

I studied mathematics at Cornell, with a concentration in mathematical physics, and played flute in the Cornell Symphony. I trained classically, played at a competitive level in my youth, and still play nearly every day. I also enjoy soccer, electronics, and 3D printing. More recently, I’ve become interested in FPV drones, particularly the possibilities for offboard vision and multimodal models.