Developing an autonomous underwater robot for hull cleaning and inspection. Leading the electrical and software architecture, including embedded control, sensing, onboard compute, autonomy, and ROS 2 integration.
Designed and tested embedded control and data-acquisition hardware for CMU Racing's electric race car, including STM32-based PCBs, vehicle networking, sensing, and real-time firmware.
Designed and optimized branch predictors under a 64 KiB hardware budget, exploring TAGE-SC, TAGE-NN, and TAGE-NN-L architectures with statistical and neural correctors on SPEC workloads.
Designed a 7-stage RV32I processor with pipelined execution, caching, hazard handling, and adaptive branch prediction, then synthesized and evaluated the design for performance.
Built a preemptive embedded operating system for STM32 with rate-monotonic scheduling, task admission control, execution-time enforcement, and mutexes with priority-inversion protection.
Implemented visual tracking, feature matching, augmented reality, epipolar geometry, 3D reconstruction, neural networks, and prompt-based segmentation culminating in 3D object point-cloud reconstruction.
Building a custom smartwatch with activity and biometric sensing, environmental sensors, gesture-based wake detection, and Bluetooth Low Energy connectivity on custom embedded hardware.
Explored reinforcement-learning agents operating under practical memory, numerical-precision, and inference-latency constraints, evaluating the tradeoffs between resource usage and policy performance.
Built an interactive alarm clock that requires completing a cognitive or physical challenge before it can be dismissed, including memory games and vision-based exercise recognition.
Implemented Pong entirely in digital hardware, including game logic, real-time video generation, user input, and a VGA display interface.
Developed labs and starter software for a CMU robotics course covering sensing, controls, actuation, computer vision, and autonomous robot navigation.