Selected work

Robotics systems, built for the real world.

Selected engineering work across wearable robotics, field sensing, embedded systems, and force-aware automation.

  • Robotics software
  • Computer vision + ML
  • Controls + automation
  • Embedded sensing
  • Physical prototyping
  • Data + validation
  • Human-centered interfaces
Aerial agricultural field imagery showing the kind of real-world setting PerPlant works in

02 · Agtech robotics · sensing + data quality

Robotics co-op · 2025

PerPlant

Integrated thermal and GPS sensing, then curated representative samples from a large field-image dataset.

Fleet Robotics underwater robots against a red wall

03 · Underwater robotics · interfaces + embedded systems

Robotics co-op · 2022–23

Harvard Microrobotics

Built fleet-operation interfaces and supported embedded communication and drive-module integration.

Completed Teradyne Cartesian robotic test system installed over the connector fixture

04 · Mechatronics · closed-loop automation

Senior project · 2023

Teradyne robotic test system

Programmed force-aware connector mating and automatic tool changes for a published robotic test system.

Independent products

The same systems thinking, applied to everyday experiences.

SunnySips forecasting and recommended visit-times interface

05 · Independent software · real-world recommendations

Released independent product

SunnySips

Released an iOS and web product that turns environmental modeling into useful outdoor recommendations.

TRYBE Boulders Sessions mobile interface

06 · Independent prototype · participation systems

Independent product prototype

TRYBE

Prototyped beginner-friendly sessions, partner perks, and booking states for real-world movement.

Academic explorations

Coursework that extends the systems story.

Relevant technical studies, kept visibly separate from professional and flagship project work.

Poster showing autonomy sensor-data and mapping coursework

DTU coursework · 2025

Autonomy, sensing, and mapping

Robotics coursework across ROS2 occupancy-grid mapping, LiDAR and GNSS/IMU data, point-cloud workflows, and Crazyflie / OptiTrack controller validation.

  • ROS2
  • LiDAR + GNSS/IMU
  • Occupancy grids
  • Point clouds
View poster
Poster showing the Super Mario diffusion coursework project

Advanced deep learning coursework · 2025

Discrete diffusion for Super Mario levels

A PyTorch U-Net diffusion study using one-hot tile representations, with attention to why image metrics alone do not capture structural validity or playability.

  • PyTorch
  • U-Net
  • DDPM
  • Model evaluation
View poster