Sean Yang

Software Engineering @ University of Waterloo, class of 2030

Waterloo, ON

Spidey & Armie

3d twin that steers a real robot

A live 3D reconstruction of a robot table: click anywhere and the real four-legged robot walks there, and when obstacles block every path, a robot arm carries it past. Built at Hack the North 2026 with three teammates. I built the tracking and navigation that keep it in sync: a C++ camera tracker on a QNX Raspberry Pi, MobileSAM object detection, 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, an EKF-tracked shuttle, and the same arm URDF the hardware runs. Trained a privileged-state PPO teacher, distilled it into a student on noisy perception, then fine-tuned the student with an asymmetric actor-critic, taking hits 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

Architected an end-to-end C++/ROS 2 stack for autonomous agricultural rovers, owning sensor integration, planning, and control across two robot platforms. RTK-odometry fusion and microsecond-level sensor sync into real-time SLAM held mean follow error under 20cm over 500m of autonomous tests.

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 built a Rust generator that makes labelled training data instead. It grows procedural apple trees through a PyO3 bridge to The Grove, plants them in rows on sculpted terrain with trellises and ground cover, and simulates the LiDAR's returns at 4M+ points per second. Every point carries a tree, trunk or background label and an instance ID, and each block ships 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 vehicle built by Waterloo's student design team. Rapid-inference models predicting how external agents behave cut safety incidents by 3x. Integrated those models with the perception, world modelling, planning, and action nodes, and maintained Docker environments to keep the stack reproducible.

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

watonomous.ca

swarm_ws

robotic swarm maze exploration

Four robots map procedurally-generated mazes with no central coordinator, exploring 7.5x faster than one robot alone. A geodesic Voronoi partition over the merged map gives each robot the region it can reach fastest, so they spread out instead of racing for the same frontier. Four ROS 2 packages with per-robot state machines run across seven Docker services, with Gazebo simulation and live Foxglove visualization.

Python · ROS 2 · Docker

github.com/RRohan4/swarm_ws

Wunder RNN Challenge

top 5% market prediction model

A Transformer time-series model in PyTorch for the Wunderlab ML Challenge, predicting the next state of noisy 1000-step financial sequences. Combining sinusoidal and learnable positional encodings captured long-range dependencies better than either alone. It scored a mean R² of 0.396 against a 0.309 participant average, placing top 5% of 3000+ competitors.

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