Projects
Morphi
Python, Modal, vLLM, PyTorch, LoRA, Qwen3-VL, React Native
Morphi is an AI-powered sports coaching mobile app built with React Native that provides real-time, precision biomechanical feedback via voice narration and visual overlays.
For this project, I architected the overall backend data pipeline and authored a custom data processing script to curate training data. I utilized Gemini 3.0 Flash Thinking to distill over 2,300 labeled examples for fine-tuning a 32B vision-language model via LoRA on 8x H200 GPUs.
To support this at scale, I engineered our serverless backend infrastructure on Modal, optimizing multi-modal inference using vLLM across a 4-GPU cluster. This architecture reduced full video processing and inference latency from 3 minutes down to under 24 seconds.
Custom CoreXY 3D Printer
Designed and constructed a high-performance CoreXY 3D printer from the ground up, featuring a rigid frame modeled in Autodesk Inventor and custom 3D-printed components.
Configured Klipper firmware on a Raspberry Pi 4 to drive salvaged Ender 3 electronics, achieving high-speed printing at over 1500 mm/s with 7000 mm/s² acceleration. The video above shows the printer printing a benchi.
W.A.N.D.E.R. Autonomous Robot
Engineered a maze-solving autonomous robot featuring a custom 3D-printed chassis, integrating a gyroscope, wheel encoders, and distance sensors for robust localization.
The video above shows the robot following a preprogrammed path through boxes. Implemented a C++ pathfinding and SLAM algorithm on an ESP32 microcontroller to efficiently map and navigate unknown maze environments.