Building AI systems that understand humans
I work at the intersection of human data and machine intelligence. Currently building human-AI collaboration at Fleet AI.
I work at the intersection of human data and machine intelligence. Currently building human-AI collaboration at Fleet AI.
A timeline of where I've been and where I'm heading
Human-AI collaboration at scale. Building training gyms for agents—high-fidelity simulations where AI practices tasks and humans provide supervision.
Shaped how frontier models learn from human feedback. Designed data collection strategies and posttraining pipelines in Meta's skunkworks for superintelligence.
Data strategy for FAIR's agent research. Scaled evaluation pipelines and training environments for agentic AI systems.
Scaled code generation and AI safety data pipelines. Managed vendor relationships. Promoted to team lead within 6 months.
Studying how blood vessels remember injury at the molecular level. Building computational pipelines for multi-omics data to uncover epigenetic mechanisms in endothelial dysfunction.
Built automated systems for e-commerce arbitrage. Scaled to six figures by 16, funding my way through college. A decade of internet hustle.
Started a nonprofit that served 50K+ people across New York's Capital Region. Paused operations after relocating. 518cares.org
Summa Cum Laude. NSF research fellow, built open-source ML tools, and somehow still found time to run a business on the side.
NSF Grant #1831547. Developed PySpady, an open-source Python library implementing sparse dictionary encoding for spatiotemporal tensor decomposition. Applied to geospatial and time-series datasets at scale.
Built internal tools and access management systems used by 4K+ employees. Python, Flask, and enterprise data infrastructure.
Most people find systems thinking in a classroom. I found it at 16, building bots to automate sneaker reselling and scaling that to six figures before college. That scrappiness led me to computer science, then to Meta, and now to Fleet.
At Fleet AI, I'm building training gyms for AI agents—high-fidelity simulations where AI learns to use real software and humans provide the supervision. The bet: drop-in virtual assistants that let anyone direct thousands of agents in parallel, accelerating the shift from performing work to directing it.
I also founded 518 Cares, a nonprofit serving the Capital Region—because building things that matter isn't limited to code.
Listed alphabetically, not by importance
Nonprofit organization focused on community support and social impact initiatives in the Capital Region of New York.
Building executable eval rubrics from real expert workflows and process-linked reward models for RLHF training data. Client: [redacted]
Autonomous 24/7 cryptocurrency trading bot operating on the Aster DEX, powered by Grok API and running on a Linux server in Amsterdam.
iOS personal trainer using computer vision to detect jump shot form and provide real-time voice feedback for improvement suggestions.
College football AI data layer serving as a coaching assistant to professionals, designed to eliminate the need for traditional data analysts.
Dual-strategy system for prediction markets: Python sentiment analysis for directional trades + Rust HFT engine for arbitrage.
Bayesian optimization techniques applied to Premier League teams and players to analyze and predict hypothetical match scenarios.
Sparse dictionary encoding Python library developed as an extension of several machine learning research papers.
I'm always building the next thing.
If you're working on something at the frontier of AI and human collaboration, I'd like to hear about it. The best conversations start with a simple message.