Where software meets the physical world

I work across AI systems, robotics, and embedded intelligence — the layer where software meets the physical world.

Most of my work lives at that intersection. At Santa Clara University's Wireless Intelligent Networks (WIN) Lab, I research FPGA-based acceleration for machine learning and wireless-communication algorithms — training and quantizing models in PyTorch and TensorFlow, then prototyping and optimizing fixed-point implementations of signal processing and channel estimation for high-throughput, low-latency workloads. That work sits close to the metal, where the constraints of real hardware shape what an algorithm can actually do.

Robotics is the other half. As a robotics hardware and embedded-systems intern at Rainier Labs, I worked on an expressive robotic head — integrating displays into a facial-expression system, redesigning CAD models so hardware fit cleanly, and evaluating audio components for the build. Earlier, with a student team, I co-developed BALANCE: a hybrid legged-and-wheeled mobility device designed to support people across the full continuum of mobility, rather than forcing the binary choice most assistive devices impose. The system was filed as a provisional patent, and a related article about the design was published in the Youth Innovation Journal.

I also build fast. At hackathons I prototype applied-AI and voice-agent products — agentic arrival orchestration, voice-first accessibility tools, native AI interfaces, and decision-support assistants — usually shipping a working demo in a day or two. I care about the same thing in a weekend build that I do in research: systems that hold up when real people use them.

I'm studying Applied Mathematics at UC Berkeley, and I trained in machine learning for autonomous robotics through Berkeley's ROAR Academy under Prof. Allen Yang. I'm most interested in problems where intelligence has to run under real constraints — limited compute, real sensors, real environments — and in building the systems that make that work.