I build things that help people — medical devices calibrated for underserved patients, assistive technology for wheelchair users, autonomous robots that adapt to their environment. I work across the full stack: CAD to code, fabrication bench to clinical testing. Rising junior at Duke Pratt, seeking summer 2026 engineering internship.
My engineering instinct is to start with the person on the other end. A jaundice detector that misreads darker skin isn't a working device. A cryo-compression glove that takes four minutes to put on won't get used. An assistive mount that breaks after two weeks helps no one. I build with that constraint front and center — technical rigor in service of real-world reliability.
At Duke's Pratt School I'm pursuing mechanical engineering with a computer science minor and a certificate in robotics and automation — a combination that lets me move fluidly between hardware and software. In the General Robotics Lab, I'm contributing to a modular robot that physically reconfigures itself mid-mission, transitioning from aerial drone to ground rover. That kind of problem — where the mechanical design and the autonomy stack have to co-evolve — is where I do my sharpest thinking.
I'm looking for summer 2026 opportunities in robotics, medical devices, or engineering systems — teams where hands-on prototyping and software integration both matter, and where the work has a clear purpose beyond the work itself.
Transcutaneous bilirubinometers — the standard tool for detecting neonatal jaundice — are calibrated for lighter skin tones. In Uganda, where the majority of infants have darker skin, this leads to systematic underdiagnosis and preventable brain damage from untreated hyperbilirubinemia. Bilibright addresses this directly.
Our team developed a low-cost handheld device using RGBC color sensors calibrated through regression models built from phantom skin testing across a full range of skin tones. The device converts raw sensor readings to bilirubin levels on a small OLED screen. The casing is fully 3D-printed and Arduino-driven — designed to be manufacturable and repairable in a low-resource hospital environment. We're currently iterating toward a nurse-friendly interface with selectable patient settings.
Most robots are built for one environment. This one isn't. At Duke's General Robotics Lab, our team is developing a robot that takes off as a drone, lands on a mother UAV, and deploys as a ground rover — physically swapping its own modules to adapt its capabilities mid-mission. Target application: environmental surveying swarms that can reach locations no single platform can.
My contribution focuses on autonomous navigation: implementing pose estimation and SLAM using cameras and AprilTag fiducial markers in ROS2. The docking sequence — where the sub-robot must align precisely with the mother drone in flight — is where the perception and control work intersects most directly with the mechanical design.
68% of chemotherapy patients develop peripheral neuropathy — permanent numbness and pain in the hands and feet caused by nerve cell damage. Cryo-compression (simultaneous cold and pressure) is clinically proven to reduce this risk, but existing solutions are expensive, bulky, and inconvenient for clinical use.
Working with Dr. MK Anastasio at Duke Hospital, our team designed and built a cryo-compression device for patients' hands that maintains temperature between 20–27°C and pressure between 30–50 mmHg for the duration of a chemo session. The device uses thermistor-based temperature sensing and air-bladder pressure monitoring for real-time feedback — all fabricated from 3D-printed parts, laser-cut acrylic, and Arduino electronics for a target cost under $50.
For the modular robot to dock with a mother UAV, it needs to know exactly where it is relative to the docking target — and it needs to correct its orientation in real time. This project builds that capability from the ground up.
A camera detects an AprilTag fiducial marker and estimates its pose using OpenCV's solvePnP algorithm, extracting the tag's roll angle directly from the rotation matrix. A PD feedback controller then drives a servo to bring the tag to a target angle. The system closes the loop between perception and actuation — the foundation on which full docking autonomy will be built.
Off-the-shelf assistive technology rarely fits. A phone mount designed for one wheelchair joystick won't attach to another. A mouth-stick holder built for a standard chair arm won't work for a custom power chair. The gap between what exists and what a specific person needs is often just one well-designed 3D-printed part.
Through internships at ADAPT Community Network and the Westchester Institute for Human Development, I designed and fabricated custom adaptive devices for occupational therapy patients — phone mounts, head-switch adapters, mouth-stick holders, seatbelt release aids, can-tab openers. Each device started with understanding one person's specific range of motion and daily workflow, then became a CAD model, then a physical object they could use.
During long-duration spaceflight, astronauts develop internal jugular vein thrombosis at rates far above the general population — a consequence of fluid redistribution in microgravity that causes chronic venous congestion in the upper body. Current countermeasures are limited, and no mechanical solution exists that integrates with a spacesuit.
This project is designing exactly that: a wearable mechanical device integrated into the neck region of a spacesuit that applies controlled external compression to the IJV, reducing stasis without restricting mobility or neck range of motion. Three distinct design concepts are currently in low-fidelity prototyping — a lead-screw compression mechanism, a pneumatic collar, and a passive spring-loaded variant.