Biomedical engineering student on the computational track at UT Austin. I work on physiological signals and the models that interpret them, and I try to run those models on the device that recorded the signal.
Clinical AI on the phone
Stealth Clinical AI
2026An on-device language model that evaluates clinical conversations for an early-stage startup. Qwen 3.5-4B in under 500 MB of RAM.
Localabs
2026Turns a photo of a lab report into a plain-language explanation in about 12 seconds, with MedGemma 4B running on the phone.
Scribend
2026An offline medical scribe, built at the Qualcomm × Meta ExecuTorch Hackathon, that writes a SOAP note from a clinic visit on the phone.
Signals and embedded
Cloud to Edge Heart Rate Monitor
2025–26A Raspberry Pi 5 streams ECG recordings to a Pico, which classifies each heartbeat with a neural network written in C while its second core acts as a watchdog.
Longhorn Neurotech
2024–Edge AI Team Lead. Our EMG gesture model runs 2.4× faster on a Raspberry Pi Zero 2 W after INT8 quantization, at 311 KB.
Embedded Multiplayer Racer
2025A two-player racing game in bare-metal C on two TI MSPM0 boards, kept in sync over UART. Voted Best Game in ECE 319K.
ECG PCB
2025An ECG board built with two classmates that produced a clean trace in LabVIEW, using an AD620 amplifier and 0.32 to 39 Hz filtering.
Research tools
Laser Speckle Imaging
2023–25Real-time 16-bit speckle contrast over a 7×7 window, with closed-loop laser current control.
phy 3D Feature View
2025A 3D view for an open-source spike-sorting tool, built at the Recanzone Lab, UC Davis.