Current projects.
What I'm building right now.
Solaris-001
The Solaris-001 is an autonomous quadcopter drone I'm designing as a personal engineering project, with the goal of operating for extended periods without manual recharging. Onboard solar panels help harvest energy both during flight and while stationary, extending overall operating time. When the battery reaches a low charge level, the drone is designed to autonomously return to a solar-powered charging station, dock, recharge, and resume operation.
I'm also exploring computer vision and autonomous navigation so the drone can interpret its surroundings in real time, avoid obstacles, and perform simple tasks independently. Potential applications I'm considering include package delivery, environmental monitoring, and surveillance. I designed the components and full assembly in SolidWorks, including a parametric airframe and interactive exploded assembly, and then 3D-printed the custom parts for prototyping. I'm currently assembling the system and beginning initial testing and iteration in my free time.
GymBro
Research from Stanford found that people who felt they were working together persisted 48–64% longer on challenging tasks than those working independently. This is called “social synergy.” GymBro takes that principle into fitness and nutrition by addressing two major problems in one platform: consistency and accuracy.
The first is consistency. Most fitness and nutrition apps are built as individual experiences, leaving users to rely heavily on self-motivation. GymBro is designed around social accountability, allowing friends and gym partners to share progress, maintain streaks, motivate one another, and work toward their goals together. The idea is to make staying on track feel less like an isolated responsibility and more like a shared experience.
But consistency only works if tracking itself is simple and trustworthy. That leads to the second problem: nutrition-tracking accuracy and convenience. GymBro uses AI-powered food analysis to let users photograph a meal and quickly receive accurately estimated calories and macronutrients, creating a fast snap-and-go experience instead of requiring extensive manual logging.
Alongside the current platform, we are researching more specialized computer-vision methods using depth estimation, structural analysis, and other modern methods to improve accuracy, particularly for portion sizes and mixed meals. By combining social accountability with intelligent nutrition tracking, GymBro aims to make nutrition easier to track and fitness goals easier to maintain. The first version of the app is currently in development and will be coming to the App Store soon. Stay tuned!
Research & prior work.
Engineering research and competition projects completed before UC Irvine. Click any cover to read the full presentation.
Aircraft Winglet Aerodynamics & Efficiency
A wind tunnel study comparing three wingtip geometries — no winglet, sharklet, and split-scimitar — to find which best weakens the trailing vortex. I built the tunnel myself and measured vortex strength by tracking how far a tethered string deflected in the wake. The split-scimitar cut average deflection from about 1.53 to 0.98 inches, roughly a 35% reduction, confirming the hypothesis that the added lower blade further delays high-pressure air rising over the tip.
I later rebuilt the study from the ground up with 3D-printed test articles and an upgraded wind tunnel, adding machine learning image analysis to characterize vortex formation across the same three geometries.
2nd Place — Fluids & Aerodynamics, OCSEFMedical Device Research & Development
I proposed Normothermic Ex-Vivo Heart Perfusion (NEHP) as a testing platform for artificial heart valves, adapting Yale's BrainEx work on restoring cellular function after death. Circulating oxygenated blood at body temperature slows post-mortem deterioration, and pacing the heart with external electrodes at 60–100 BPM recreates realistic pressure and flow around a valve under test.
The research mapped where current validation falls short — mechanical simulators that don't reproduce true cardiac dynamics, animal studies that are costly and ethically fraught, and simulations that can't yet model full biological complexity — and argued a perfused heart offers better fidelity while reducing live-animal use. The work was guided by an R&D engineer at Edwards Lifesciences.
Biomechanical Shape-Memory Alloy Actuation
A tendon-driven prosthetic hand actuated by Nitinol shape-memory alloy springs instead of motors. Running current through a spring heats it past its transition temperature into the austenite phase, contracting it and curling the fingers; cutting the current lets it cool back to martensite and relax — the same role the forearm flexors play in a real hand.
Shape-setting the wire and tuning that transition temperature above ambient took repeated trial and error, and a single shared spring couldn't overcome the rubber-band extensors, so I moved to an individual spring per finger. The final build articulated every finger on demand.