Robotics M.S. student at Georgia Tech. I build embedded control systems and robot software, from STM32 boards to learned policies.
Led early avionics architecture for a student VTVL rocket, coordinating with guidance, propulsion, and mechanical teams to define subsystem responsibilities and electrical interfaces.
Independently designed and laid out a multipurpose STM32 controller PCB with CAN, UART, I²C, and SPI interfaces for propulsion and thrust-vector-control applications; teammates later validated the prototype.
Contributed to architecture design and core experiments for an imitation-learning model that uses optical flow as action primitives to learn from action-free demonstration videos; extended Meta-World and LIBERO for Sawyer and Panda arms and demonstrated cross-embodiment transfer.
Co-led a 10-member software team, owning the light-show editor and backend and coordinating their interfaces with the Raspberry Pi control team.
Rewrote the TypeScript backend in Rust using SQLx and GraphQL for typed and efficient processing of deeply nested lighting data. The new backend supported multi-user editing of scenes with 5,000+ LEDs and 1,000+ keyframes and cut per-frame saves from 2–3 seconds to a few hundred milliseconds.
Led navigation development for an autonomous robot aboard the ISS, combining circle-based target detection and ArUco pose estimates with a Dijkstra planner that routed around polygonal keepout zones.
Chose repeated pose correction and conservative motion for final target engagement; the team earned full shooting points in its final competition run and the highest overall score worldwide.
Investigated Neural Radiance Fields (NeRF) and 3D Gaussian Splatting for static and dynamic scene reconstruction, with a focus on dynamic scenes.
Independently initiated a differentiable sim-to-real RL project using JAX and NVIDIA Warp to correct simulator dynamics from target trajectories; evaluated locomotion policies under simulated dynamics shifts.
Co-built a Mobile ALOHA pipeline spanning ROS 2 data collection, camera calibration, optical-flow estimation, model training, and on-robot inference.
Enhanced LLaVA-1.5-7B for multimodal autonomous-driving reasoning using a two-stage retrieval-augmented generation (RAG) pipeline and auxiliary visual features, improving evaluation score by 21% on the CODA-LM dataset.
Designed and built the four-wheel robot's mechanical structure, power system, and automated frisbee loading and lifting mechanisms.
Built an STM32 audio pipeline using I2S, DMA, and CMSIS-DSP to compute Mel spectrograms from live performance.
Linearized a provided Furuta pendulum model, designed and tuned an LQR controller in MATLAB/Simulink, and tested upright stabilization and disturbance recovery on course hardware.
Built a BLE-based shopping-cart prototype for indoor product navigation and checkout; the integrated design received a Taiwan utility model patent.
Dec 2025 – Jul 2026
Embedded Engineering Intern
Jul 2024 – Nov 2025
Undergraduate Researcher - 3D Scene Reconstruction
Jun 2025 – Aug 2025
Software Engineering Intern - Embedded AI, Algorithm, Tools
Aug 2022 – Aug 2025
Undergraduate Researcher - Robot Learning
2023 – 2025
Teaching Assistant
Jul 2023 – Aug 2023
Algorithm Summer Intern
I studied electrical engineering at National Taiwan University, where I did robot-learning and 3D-reconstruction research and led student teams on a space-robotics challenge, a self-landing rocket and a light-dance show. I have interned at Tensor Tech, Ambarella and Academia Sinica.
Learning Skill Abstraction from Action-Free Videos via Optical Flow — ICML 2025 Workshop
Hung-Chieh Fang*, Kuo-Han Hung*, Chu-Rong Chen, Po-Jung Chou, Chun-Kai Yang, Po-Chen Ko, et al.
