> Computer vision developer on a combat-robotics student design team — building the perception and autonomy stack that finds, tracks, and targets opposing robots in real time.
// perception → actuation
Real-time C++ detection system on a Jetson Orin Nano that automates aiming by tracking and predicting enemy robot motion — an OpenCV pipeline extracts precise target coordinates and angular orientation, with camera calibration tightened to cut RMS reprojection error by ~80%.
Processed over 12,000 images for the detection model, accelerating annotation by distilling labels from the team's older model and expanding the number of detected classes 4×.
Implemented motion planning on the robot stack using Manifold-tech's Odin ROS drivers and Nav2, with Foxglove Studio for live introspection and debugging.
Wrote custom ROS nodes that stream planner and targeting data over serial to the microcontroller, closing the loop between high-level autonomy and low-level actuation.