I earned my M.S. in Robotics from Northwestern University and my B.S. with honors in Mechanical Engineering with a Computer Science minor from The Ohio State University. Most recently I was a robotics software engineer at a stealth startup, shipping VLA and diffusion-policy stacks and real-time perception for dual-arm manipulation on hardware.
Fine-tuned VLA models (pi0 family) for dual xArm7 manipulation; built synthetic data pipelines with SAM2 / CUTIE; improved diffusion policies with segmentation cues; real-time perception with FoundationPose and Grounded-SAM for precise grasping.
Details
Coursework and research in SLAM, manipulation, machine learning, and computer vision.
Path planning and control for mobile robots; robotic arm design; deep learning for perception.
Teleoperation · Imitation learning · VR · Simulation
Low-cost, high-fidelity master–slave teleoperation on the OpenArm platform for imitation learning: a custom Dynamixel master and VR teleop (6‑DoF over ROS 2), synchronized RGB + joint logging, LeRobot Diffusion/VLA training, and full deploy. Isaac Sim complements hardware with VR teleoperation and dataset collection.
Imitation learning · Ergodic control · Haptic teleoperation
Leveraged ergodic imitation to learn from both positive and negative demonstrations. Built a haptic-guided teleoperation rig where one 7-DoF Franka in impedance mode is mirrored by a second Franka, then trained robust task models with the ergodic metric — a measure of information content in motion.
Impedance control · Compliant manipulation · Franka Emika
Implemented impedance control on a 7-DoF Franka Emika collaborative arm. Compliant interaction enables safe contact with humans and objects, and serves as the substrate for downstream haptic teleoperation and learning-from-demonstration work.
Extended Kalman Filter · SLAM · Simulation · C++
Built an Extended Kalman Filter SLAM stack from scratch in C++ as a set of ROS packages — turtlelib (SE(2) math), odometry, sensor models, and EKF estimation — validated in simulation on a TurtleBot navigating a landmark-rich environment.
Computer Vision · Manipulation · NP-hard
Solved the 3D bin-packing optimization problem on a Franka Emika Panda. Computer vision detects object dimension and pose; MoveIt2 plans collision-free trajectories that pack items efficiently into a container while minimizing wasted space.
Trajectory planning · Feedforward control
Trajectory planning, odometry, and feedforward + feedback control for a KUKA youBot — a mecanum-wheel mobile base with a 5R arm. The robot picks up a block at a known pose, transports it across the workspace, and places it at a target location.
Dynamics · Simulation
Simulated a jack bouncing inside a moving box from first principles. Derived configuration frames and Lagrangian dynamics, then integrated the equations of motion over a 10 s horizon at a 0.01 s timestep, capturing collision and contact behavior.
Computer vision · Manipulation · Motion planning
Used a Franka Emika Panda to brew pour-over coffee end-to-end. Computer vision with AprilTags localizes each tool; a custom MoveIt wrapper drives a multi-step recipe (grind → kettle → pour → discard). Built with a team of five over three weeks.
Computer vision · Depth sensing · Manipulation
Tracked a purple pen in 3D with an Intel RealSense depth camera and commanded an Interbotix arm via the xs toolbox to grasp it in real time. Inverse kinematics maps the live target pose to joint commands for closed-loop pursuit.
