Sean Yang

Software Engineering @ University of Waterloo, class of 2030

Waterloo, ON

Spidey & Armie

3d twin that steers a real robot

My team of four built a live 3D model of a robot table at Hack the North 2026, and clicking anywhere in it sends a real four-legged robot walking to that spot. If obstacles block every path, a robot arm picks the robot up and carries it over them. I worked on tracking and navigation: a C++ camera tracker running on a QNX Raspberry Pi, object detection with MobileSAM, and path planning.

C++ · TypeScript · OpenCV

github.com/aicheye/HTN26

Badminton Receive Policy

Robotics SWE @ WATonomous

Built the badminton receive environment in mjlab on MuJoCo Warp, with shuttle aerodynamics, EKF shuttle tracking, and the same arm URDF the real hardware uses. Trained a PPO teacher on privileged state, distilled it into a student that only sees noisy perception, and fine-tuned the student with an asymmetric actor-critic. That raised the hit rate from 76.6% to 98.8% of serves, with 78% returned over the net.

Python · PyTorch · mjlab · C++ · ROS 2

watonomous.ca

Autonomous Row Following

Perception Engineering Intern @ moss

Built a C++/ROS 2 autonomy stack for agricultural rovers end to end, covering sensor integration, planning, and control on two robot platforms. Fusing RTK with odometry and syncing sensors to the microsecond for real-time SLAM kept the mean follow error under 20cm across 500m of autonomous testing.

ROS 2 · Rust · C++ · Linux

moss.ag

Synthetic Pointcloud Pipeline

Perception Engineering Intern @ moss

Only about 50K of 3M real orchard LiDAR scans were labelled, so I wrote a Rust generator to make labelled training data instead. It grows procedural apple trees through a PyO3 bridge to The Grove, plants them in rows on shaped terrain with trellises and ground cover, and simulates LiDAR returns at over 4M points per second. Every point is labelled as tree, trunk, or background with an instance ID, and each generated block comes with a streaming COPC viewer.

ROS 2 · Rust · C++ · Linux

moss.ag

EVE Prediction Pipelines

Robotics SWE @ WATonomous

Wrote concurrent C++/ROS 2 software for perception and planning on EVE, the autonomous car built by Waterloo's student design team. Fast-inference models predicting what other road users will do cut safety incidents by 3x. Also connected those models to the perception, world modelling, planning, and action nodes, and maintained the Docker environments so the stack builds the same way everywhere.

Python · PyTorch · mjlab · C++ · ROS 2

watonomous.ca

swarm_ws

robotic swarm maze exploration

Four robots explore procedurally generated mazes together without a central coordinator, and they finish 7.5x faster than a single robot. A geodesic Voronoi partition of the shared map gives each robot the area it can reach fastest, so they don't all go after the same unexplored corner. It's four ROS 2 packages with a state machine per robot, running across seven Docker services, simulated in Gazebo and visualized live in Foxglove.

Python · ROS 2 · Docker

github.com/RRohan4/swarm_ws

Wunder RNN Challenge

top 5% market prediction model

A Transformer model in PyTorch for the Wunderlab ML Challenge that predicts the next step of noisy financial sequences 1000 steps long. Using sinusoidal and learnable positional encodings together picked up long-range patterns better than either one did alone. It scored a mean R² of 0.396 against an average of 0.309, which put it in the top 5% of more than 3000 entrants.

Python · ML · Jupyter

github.com/aicheye/wundernn

Sean Yang

Experience

Projects

Links

Sean Yang

Spidey & Armie

Badminton Receive Policy

Autonomous Row Following

Synthetic Pointcloud Pipeline

EVE Prediction Pipelines

swarm_ws

Wunder RNN Challenge