I study computer science at CMU and am an intern at TikTok.
A Rust workspace I use for market data, backtests, live trading, and strategy research.
The useful part is keeping research and live code close together: exchange clients over TCP/WebSockets, normalized data in DuckDB/PostgreSQL, Black-Scholes/SABR pricing, and Forust/ONNX models behind isolated execution and risk gates — all in one workspace.
An agentic workflow layer for AI coding harnesses — planning, implementation, review, docs, and release, all behind host-native skills.
My Rust 2D engine. I use it to learn the parts of graphics I care about: rendering, text, UI, assets, and game loops.
I'm on TikTok's Intelligent Creation team, working on effect cloud rendering — moving graphics-heavy effects off the phone and into the cloud so they run the same anywhere. Lately I've been building a Go service that runs AI image-to-image and image-to-video effects, and adding cloud support for rich text, animated subtitles, and gesture-driven effects.
Worked on the crawler, XSS scanner, secret scanner, and backend that ran scans and reported results.
Helped build a multimodal autism behavior dataset and benchmarked visual models against it. That work led to papers in IEEE T-MM and CVPR Sight and Sound.
Journal version of our work on recognizing autism-related behaviors from audio and video.
Workshop paper on using multimodal LLMs with synchronized audio and video for autism behavior recognition.
I like talking about graphics, systems, security, and trading. Email is best.
Contact me