I build applied AI systems, agent infrastructure, and data products—with a particular interest in trustworthy automation, local-first software, and financial intelligence.

I earned an M.S. in Electrical Engineering from Columbia University, where my work covered machine learning, NLP, generative AI, and large-scale data. I like turning research ideas into software people can inspect, run, and improve.

A native Rust coordination runtime for concurrent AI agents. Independent Codex, Claude Code, Cursor, and custom workers get shared durable memory, direct and broadcast messaging, PID/session discovery, wakeable event streams, and atomic task leases through one private SQLite database—without a daemon or cloud account. Its durable-write path is 11.7× faster than the original Python runtime.

Rust · MCP · Multi-agent systems · SQLite · Local-first

An evidence-gated evaluation harness for financial research agents. It catches uncited claims, look-ahead evidence, stale sources, numeric conflicts, and weak high-confidence conclusions—locally and without an API key. It ships as a Python CLI, GitHub Action, and reusable agent skill.

Python · AI evaluation · Agent skills · Fintech · Local-first

A local-first macOS portfolio risk laboratory built with SwiftUI and Swift Charts. It includes allocation analytics, reproducible Monte Carlo simulations, interactive stress testing, and validated CSV import—without uploading financial data.

Swift · SwiftUI · Swift Charts · Fintech · On-device research

A zero-runtime-dependency Python CLI and GitHub Action that converts repository metadata into an actionable, transparent community-health report. It supports Markdown, machine-readable JSON, and live Shields endpoint badges.

Python · GitHub API · GitHub Actions · Developer tools

Research and experiments around fine-tuning language models for medical-domain workflows.

LLM fine-tuning · Healthcare AI · Applied research

• Reliable agent harnesses: evaluations, guardrails, observability, and recovery
• Process-aware local agent coordination, shared memory, and safe work handoffs
• On-device intelligence with Core ML and privacy-preserving local workflows
• Explainable portfolio analytics and probabilistic financial risk models
• Retrieval systems and multimodal embeddings grounded in domain knowledge

I'm happy to collaborate on substantive open-source work in AI infrastructure, fintech, macOS, and applied ML. Open an issue on one of the projects above or connect with me on LinkedIn.