Robotics Engineer
Building systems that perceive, learn, and move.
M.S.E. Robotics · University of Pennsylvania
Blog & paper collectionReading notes, concepts, and writingI'm an M.S.E. Robotics student at the University of Pennsylvania, working on robot learning and autonomy. My research spans generative world models with Jiatao Gu's group, image-based policy learning at xLAB, and perception and control for HAMR at ModLab.
I'm currently interested in using in-context learning and world models to build robotic policies that adapt to new tasks and environments.
Researching generative world models and conditional image generation, alongside in-context learning and on-policy distillation for robotic policies.
Researching learning signals for image-editing policies through conditional generative modeling. Developing robot policies by minimizing differences between latent representations of current and goal images.
Developing HAMR perception and control tools, including vision-based motion estimation of a custom ball caster, visual SLAM, localization, sensor integration, and PID control. Developed a simulator and path-planning algorithms to study mobile robot behavior.
Implemented autonomous yard patrols using a quadruped robot, enabling waypoint navigation, dynamic obstacle avoidance, sensor-based image capture, and web-based monitoring and control. Created full-stack AI guest check-in system, supporting visitor/employee appointments and real-time status tracking.
Developed a Python web scraper using Scrapy to extract and process over 100 million data hashes, storing results in PostgreSQL. Authored a software requirements specification outlining system functionality, technical requirements, and project scope.
Robot learning · Sim-to-real
Teaching a drone to race three laps with PPO. From shaping a power loop to letting the policy find a faster line, then preparing it for real flight with domain randomization.
ModLab · Mobile robotics
Connecting caster mechanics, visual sensing, and feedback control in a holonomic mobile robot. Work spans dual-camera caster tracking, depth-camera localization, and ROS 2 / embedded PID control.
Robot learning · Reinforcement learning
A residual reinforcement learning framework around a frozen vision-language-action policy. SAC and PPO learn bounded action corrections through a shared training and evaluation pipeline.