As an undergraduate researcher in the Adie Lab at Cornell, I expedited rapid testing and iteration of an Acoustic Radiation Force - Optical Coherence Tomography (ARF-OCE) system by designing a custom waterbath fixture to align ultrasound and OCT imaging systems in a tissue-mimicking setup and spearheaded the integration of finite element modeling simulations to the lab's design cycle.

• Designed and fabricated the fixture aligning optical and acoustic paths
• Improved resolution of the ARF focal spot with a newly mounted transducer
• Iterated CAD and tested mechanical alignments to integrate both transducers
• Spearheaded finite element modeling to reconstruct phantom mechanics and predict ARF effects
• Presented in two lab meetings and submitted a final research report

As part of Robotics Studio at Columbia, we were challenged to design, build, and program a fast-walking robot using only 8 servo motors, materials available at home or in the Makerspace, and a strict $100 budget. I created Rainbow RoboCat — a quadruped robot with serial-driven legs, a rainbow-gradient 3D-printed chassis, and playful feline features.

From early concept sketches to a detailed CAD assembly, I engineered every component myself - modeling every nut, bolt, wire, and joint in Autodesk Inventor before moving to physical fabrication. I printed all chassis components on my at-home Ender 3 V3 SE using silk rainbow PLA, sanding and finishing each part to give RoboCat a polished look. I also learned how to solder for the first time, wiring the LX16A servos, battery, Raspberry Pi, and other embedded electronics into a compact and durable onboard setup.

Starting from a single leg, I gradually expanded control to all four legs using Python and the pylx16a library, developing motion primitives and a working sinusoidal gait. To further boost RoboCat's speed, I built a full PyBullet simulation of the robot using a custom URDF and applied a hill climber and genetic algorithm to optimize the gait's sinusoidal parameters.

The result: a robot capable of reaching a top speed of 38 cm/s, with movement tuned both in simulation and validated on hardware — all while staying under budget, finishing within one semester, and keeping Rainbow RoboCat full of personality.

Over my three years on Cornell Electric Vehicles, I grew from a general mechanical subteam member into Drivetrain Subsystem Lead, a role I held for about six months. During this time, I led a dedicated assembly and jigging team of over a dozen people through the most precision-critical phase of the car's build: physically integrating the drivetrain and brake subsystems into the carbon fiber chassis.

I coordinated closely with other subsystem leads while personally designing and iterating multiple custom drilling jigs in Autodesk Inventor. These combined 3D-printed stencils, laser-cut acrylic, and 8020 framing to align drivetrain components to CAD-based constraints, all while navigating real-world tolerances and foam-core crush risks. I developed the wheel well and axle alignment jigs and directly participated in shop fabrication, using mill, drill press, and laser cutter tools comfortably to ensure fast turnarounds.

One of our initial floor jigs failed due to an oversight in the support geometry. Thanks to quick thinking from teammates I was managing, we pivoted to a laser-cut acrylic solution and salvaged the drilling schedule just in time. I learned to manage shifting priorities, communicate intent across teams, and respond constructively to mistakes under time pressure.

Later, I helped resolve unexpected interference between drivetrain and steering components by co-designing a machined brake mount fixture on short notice. These experiences sharpened my skills in DFM, cross-functional collaboration, and real-world, edge case mechanical problem-solving. The drivetrain's successful competition debut was the product of intensive planning, close communication, and a lot of hands-on shop work.

This project collection explores the cardiovascular system through fluid dynamics simulation, numerical transport modeling, and developmental biomechanics analysis.

In a computational fluid dynamics (CFD) experiment using Ansys Fluent, I simulated blood flow through a bifurcating carotid artery with and without obstruction. Using realistic CAD geometries and mesh refinement strategies, I analyzed how arterial narrowing affects flow behavior — specifically, increases in wall shear stress and flow recirculation zones that can contribute to conditions like aneurysms. The results demonstrated the role of CFD in assessing cardiovascular health risks and optimizing treatment designs.

In a separate biomechanics research project, I authored a review paper on cardiac looping morphogenesis — the earliest symmetry-breaking event in heart development. I compared two major modeling approaches: an integrated finite element model based on experimental observations (Shi et al., 2014), and a geometric instability model driven by residual stress (Bevilacqua et al., 2021). This work deepened my understanding of how mechanical forces and material properties drive tissue morphogenesis and shape organ development.

Additionally, in a third project, I explored 1D and 2D transport phenomena using Python-based numerical methods to simulate diffusion and advection in blood vessels. While an exploratory project, it helped solidify my understanding of discretization schemes and the interplay between transport mechanics and boundary conditions — foundational for modeling drug delivery and solute transport in biomedical contexts.

For Columbia's Human-Centered Design & Innovation course, my team and I designed Atelier Vintage — an online marketplace that helps vintage shoppers confidently buy, tailor, style, and care for 20–100-year-old garments without the usual uncertainty or cost barriers.

Overview

We started with the hypothesis that shoppers love vintage for sustainability and individuality, but avoid wearing it because they don't know how to size, alter, clean, or style older garments. Interviews with shoppers, stylists, and sustainability advocates confirmed that vintage pieces often feel high-risk and low-support across fit, care, and styling.

My Role

I contributed to defining the core problem, synthesizing user insights, shaping the Smart-Fit and Digital Passport concepts, comparing concept directions, and developing the prototype and final narrative across pitch and demo materials.

How We Reached the Solution

We explored three concepts — an app, an online marketplace, and an in-person one-stop shop. The end-to-end marketplace emerged as the strongest direction because it addressed user anxiety around fit, care, and authenticity while remaining operationally realistic.

The Final Solution

• AI Smart-Fit for measurement capture and digital alteration maps
• Certified tailor network offering accessible, consistent alterations
• Digital Product Passport with history, care guidance, and styling inspiration
• Marketplace model with optional styling add-ons and transparent pricing

Why It Solves the Problem

The system reduces cognitive load, removes technical barriers around fit and garment care, and increases trust by making a garment's history and maintenance guidance visible — addressing core problems surfaced in research.

Challenges & Unpursued Directions

Alternative directions — like rental models, hybrid stylist-tailor apps, and immersive in-person vintage stores — were compelling but not pursued due to operational complexity or limited user pull.

User & Business Impact

Shoppers gain confidence wearing vintage pieces; platforms reduce returns tied to sizing; and tailors/stylists benefit from more predictable digital-to-physical demand.

What I Learned

This project strengthened my ability to translate ambiguous user hesitations into actionable design requirements, evaluate solution paths objectively, and articulate a full service ecosystem linking UX flows, business operations, and user value.