PROJECTS
PROJECTS
Here is a compilation of some projects I've enjoyed working on. They are taken from a mixture of work, personal projects, research, and old coursework. If you’re interested in getting to know more about any specific one, and the skills I learned from it, please get in touch.
MOMODEX
Momodex is a full-stack web app built with a 5-person team in 2 weeks. It allows users to capture images of flora and fauna, and the app identifies the species and creates custom playing cards that can be collected or used to battle the computer. The project includes integrated AI image recognition (Gemini API), public conservation data (iNaturalist API), cloud image hosting (Cloudinary), authentication (Auth0), a large database of species and playing-card data, and a custom turn-based game engine with type-effectiveness and status-effect mechanics. The back-end is built with Node, Express, and SQLite with Knex and the front-end with React. My main contributions were setting up the image to AI identification to card creation pipeline, and generating the map from public API data. I built these features from back end to front end, handling database updates, API creation, CRUD operations, and React components and routing. I also worked on some battle mechanics, upload form on the homepage, generating the card deck, and deleting a card.
Tech: React, Node.js, Express, Knex, SQLite, Google Gemini API, iNaturalist API, Cloudinary, Auth0.
See the full presentation on youtube here
See it on Github
This is a deliberately silly website all about canned SPAM, built as a team project at Dev Academy. It has mini-games (Spam Jump, Whack A Spam, Snake), a flavour rating system, comments, and a personality quiz. The goal was to practice collaborative development and keep the code quality high on a fun, full-featured app. My contribution was setting up user login with Auth0 and connecting Cloudinary for image hosting. I also designed the database and seed data for the gallery, and wrote the CRUD operations that fetch data and display it in React for the About and Gallery pages. This project shows I can take a feature from end to end: authentication, third-party media storage, database design, API routes, and the React UI that uses them.
Tech: React, TypeScript, Vite, TanStack Query, Tailwind CSS, Material UI, Node.js, Express, Knex, SQLite3, Auth0, Cloudinary, Vitest.
See it on Github
This project is currently a work in progress. I train on a Kilter Board, an LED climbing wall with thousands of routes made by the community. Users connect the Kilter Board app to their climbing wall, and by clicking on a route they can have the holds light up to show where to place your hands and feet. .I wanted to see if a machine learning model could learn what makes a good route and create brand-new climbs at a chosen grade. It also checks that every generated climb is reachable for someone of my height and wingspan. This is a work in progress. It is mainly just to help my learn python and more about ML models.
It downloads 100k+ existing climbs and turns each route's holds, positions and grade into training data. It then trains an LSTM that predicts the next hold from the holds so far and the target grade. Using that model, it generates new routes one hold at a time and draws them in the terminal as ASCII art, but could eventually output something prettier once I move to a better UI.
Tech: Python, PyTorch, pandas, Typer, Rich, pytest, boardlib.
See it on Github
BETABASE
BetaBase is a work in progress. The goal is to create a website for bouldering women and short people to share tips and tricks on how they manage to climb boulders that would otherwise be height restrictive without some creative maneuvers, or "beta". The site will mimic social media to encourage people to post videos, ask for help, and search for climbs that they might want to try next. I want to create an engaged community of support and excitement over this beloved sport, in a women-led space. It will be both functional and enjoyable, hopefully allowing more women to get involved in climbing and removing some of the existing barriers.
Tech: React, TypeScript, Node.js, Express, Knex, SQLite.
See it on Github
INNOVATION DAY PROJECT: MOTORIZED DISCECTOMY DEVICE
Innovation day is a Stryker funded project chosen and managed by employee teams. Our team of 3 designed a motorized tool to cut and evacuate intervertebral disc material from the spine for use in spinal disc removal surgeries. Current manual discectomies are done with rongeurs. The surgeon will grab and rip disc material with the rongeurs, then bring them out of the surgical site to empty them of material. This process will be repeated 20 to 80 times before all material is removed. Each time the rongeurs pass by a nerve root that if damaged, could cause numbness or paralysis to extremities. The proposed design would be motorized to effectively cut the tough material and evacuate it up a rotating auger, complete with suction and irrigation, all while fitting within the 8mm disc space. This eliminates the need for multiple passes in and out of the surgical site. This lower risk, more efficient method would benefit both surgeons and patients. Our team won 3rd place for this design
CREATING A BLUETOOTH ENABLED CLIMBING TRAINING BOARD
This was a fun personal project for me. After building a home climbing wall in my garage (see project below), I wanted to connect it to the app that would display different routes by lighting up LEDs under the holds. I installed LED strips in the climbing board, and set up an Arduino to connect to the app via BlueTooth and to display the correct LEDs on the grid when commanded. Since I'm just a novice when it comes to coding, I found a project on Git Hub and was able to modify it to work with my setup with just some minor alterations which I did in C++. Now different routes can be selected on the app, and the corresponding LEDs will light up on the board. The different colors signify start holds, middle, and finish holds.
DAISY CHAINED ENCODERS PROJECT
This is a project I did for Stryker, and I legally can't say too much about it's purpose and details. It involves daisy chaining high accuracy encoders and slip rings in a mechanically stiff but free-moving chain assembly. This one has taken many iterations to get the golden ratio of small size to low compliance. It has also required me to learn more about communication protocols and calibration methods. My role was mechanical design and project lead, and I worked closely with our electrical and controls engineers to build high-fidelity prototypes and reduce technical risks.
ROBOTIC ACTUATOR DESIGN
This was a project to develop a high-torque backdrivable actuator for robotic systems. Robots often have to have high torque outputs to be able to move heavy things slowly and carefully, but this raises safety concerns. Allowing arms to be backdrivable means that the user can push on the joint and it will move out of the way. This project utilized a state of the art gearbox and it involved determining the required torque output, accuracy, and other requirements, translating that into motor, gearbox, and encoder specs, sourcing components, designing packaging, building and testing the actuator.
ICRA 2024 ATTENDEE
I had the opportunity to attend the International Conference on Robotics and Automation in 2024. There was an amazing workshop on tracking systems which I attended, not to mention the tech expo which displayed emerging robotic technologies and precision component products. It was an inspiring experience and I left with many new ideas for future projects for my team to work on, as well as new technology options to integrate into our current projects. As the only representative from my company in attendance, I put together and educational presentation to share with the rest of my team
HIGH TORQUE LOCKING MECHANICAL JOINT
There is a patent in process for this design. It is a purely mechanical locking joint, which utilizes the principle of one-way clutch bearings. By using spring loaded rollers in opposite directions and a cable-actuated pin to disengage the locking rollers in the unlocked state, I created a joint with infinite locking positions, extremely high locking torque, and small form factor. The intended use for this device was a purely mechanical locking arm to assist surgeons in holding retractors in place. This would require a small arm with high locking torque, where all the joints could be locked simultaneously via a mechanical trigger mechanism.
KNEE LIGAMENT MODEL
The knee has ligaments that maintain stability in the joint, which is something surgeons have to consider when doing knee replacement surgeries, to ensure that the resulting implant doesn't feel too tight or too loose. I was tasked with building a repeatable and predictable model of the knee ligament interactions for training purposes and device testing purposes. The mechanism inside models the change in ligament stiffness as the joint is distracted. The device shown was assembled to a a realistic leg model and tested using force sensors, position trackers.
ICRA 2022 ATTENDEE
I had the opportunity to attend the International Conference on Robotics and Automation in 2022. There were some incredible interactive exhibitions, which meant I got to test run an exoskeleton robot (see photo). The exhibition showed some inspiring technological breakthroughs from startups and research labs across all industries, as well as some novel product releases from motor, gearbox, and sensor manufacturers. I attended a workshop on Endoluminal Intervention where the focus was on surgical robotic solutions. As the sole attendee for my company, I presented my learnings to my team
BUILDING A HOME CLIMBING WALL
This was a project inspired by my frustration with my local climbing gym. I decided to create a bouldering training wall in my garage so I planned it out and constructed it myself. This was by far the largest construction project I had taken on. Before this I had build a standalone outdoor climbing wall as a covid project, and since then I built a picnic table for my patio. This wall is integrated into the structure of the garage so it is very stable, and required some extra considerations and planning. The walls are at 40 degrees and 30 degrees for extra climbing space and difficulty, and the padding underneath is old mattresses for safety. I've been loving my new training area and have new plans all the time for improvements and additions I can make to it.
.jpg)
DESIGNING A FRACTURE FIXATION DEVICE FOR SMALL BONES
This was a project from my previous employment, while I was pursuing my Master's degree. It is a method of fixing bone fractures in fingers. Commonly known as "boxer's fracture," a break in the proximal phalanges can be hard to fix in place for proper healing. fractures are usually fixed with plates that sit on the outside of the bone, but in this case there's not much space and there is the possibility of irritating the tendons that run along the fingers. My design was for a fixation device that would sit inside the intramedullary canal, fixing the fracture from the inside. Placing the device in the canal required some bend-ability to the rod, and then the screws would fix it to the cortices of the bone.
BUILDING A REALISTIC HAND MODEL FOR CARPAL TUNNEL DEVICE TESTING
This project was to create a realistic hand model for device testing and surgeon training for endoscopic carpal tunnel release. I made a mold of my own hand using a silicone material and placed a more fibrous strip of material inside to represent the carpal ligament. the device must then be inserted through the incision and the blade extends to slice the ligament, all guided by the integrated endoscope. afterward the model can be opened and the ligament replaced with a new one. This was just a lot of fun to work on. I got to play around with some new materials and build something quite unique.
DESIGNING A RETRACTING BLADE CARPAL TUNNEL DEVICE
This device is meant for minimally invasive endoscopic carpal tunnel release procedures. It consists of an endoscope, covered by a housing which holds a retractable blade. The surgeon inserts the device through a small incision, positions it using the endoscope, then extends the blade to cut the ligament. The majority of the challenge was in designing an effective blade shape. The blade needed to extend and retract without wavering, and cut smoothly when pulled through the ligament. Each iteration was tested on cadavers and feedback was collected from surgeons.
BULDING AND PROGRAMMING A DRAWING ROBOT
This project was part of an Intro to Robotics course taught by Dr Weizhao Zhao. First, a 6DOF robotic arm was built from k'nex pieces that would mount to a table and hold a pencil. A script was written in matlab to move the end effector through various straight lines to draw a path out of a maze.
EFFECTS OF REUSING SINGLE-USE CARPAL TUNNEL BLADE CARTRIGES IN SURGICAL CENTERS
This study was a done in collaboration with the Scanning Electron Microscopy Lab at my University. I had taken a course on how to use the SEM and offered to initiate this project with my employer. The surgeon at my company had mentioned that single use medical devices are sometimes re-used multiple times at hospitals and surgery centers, often to the detriment of patients. The devices are cleaned and re-used, but may be accumulating unseen damage and biological debris with each use. I wanted to investigate the effects of this practice. This study focused on blade cartridges used in carpal tunnel release surgery. 6 cartridges were collected from a surgery center and the blades were analyzed under scanning electron microscopy and observations were recorded. 5 out of the 6 blades showed damage to the blade tip, dulling of the blade edge, and attached tissue fibers and other biological debris
CULTURING AND ENCAPSULATING PANCREATIC ISLET CELLS
While at the Diabetes Research Group, I worked on optimizing parameters for encapsulating islet cells. Mouse islet cells were cultured and fed through an encapsulation machine that would isolate and coat each to prevent immune detection. The fluid flow parameters controlled the separation and coating of cells. The resulting solution was imaged and those images were fed into a program to quantify how many were fully encapsulated.
CREATION OF A 3D BONE MODEL FROM 2D IMAGE SETS
This project was part of a biomechanics course. Images of the head of a femur bone were taken with a digital camera at recorded positions in space. A MatLab script was then created to convert the images to a 3D model of the bone surface.
SOLVING THE “JUMP ALL BUT ONE” GAME USING BLIND SEARCH ALGORITHMS
This project was part of an Intro to Artificial Intelligence course at the University of Miami. The "Jump All But One" game is played with golf tees on a triangular wooden board (most often found at Cracker Barrel or your grandmother's house). My team wrote a program to play the game using various blind search algorithms to determine which method was most efficient. Methods used in this experiment were Breadth-First Search, Depth-First Search, and Iterative Deepening. Results were compared using time to solve the puzzle and number of solutions found. Results showed Depth-First Search and Iterative Deepening were equally efficient. This is due to the nature of the game which requires a set amount of moves to win, limiting the effectiveness of better search algorithms.
BUILDING A THERMOMETER FROM SCRATCH
This project was part of a Biomedical Measurements Course. My teammate and I made a thermometer using a thermistor and a hand-built analog to digital converting circuit, a linearizing bridge, and two op-amps. The system was made for a range of 30 to 60 degrees Celcius. We first built the linearizing brige and zeroed the voltage output at thirty degrees using a variable resister. Next we designed a differentiating op amp that amplified the signal and then sent it to a summing op amp so that the voltage between 30 to 60 degrees Celcius would be .03 and .06 volts. We then constructed an AD converter with a display unit that would display these voltage values. To test our project we immersed the thermistor in a beaker of water at thirty degrees on a hot plate and increased the temperature slowly to 60 degrees recording the displayed temperature at every 10 degree increment. We repeated this three times and found the error values. Our results showed that the thermometer had a very low percent error but was found to sometimes be off by about 2 degrees which is not desirable for a thermometer, but was still an educational exercise.
WRITING A GRANT PROPOSAL FOR A DRUG-ELUTING HYDROGEL FOR BREAST CANCER TREATMENT
The purpose of this project was to practice writing grant proposals. The planned experiment was never performed, but I really enjoyed doing the background research and planning and do believe it could even work if the experiments were to be performed. Though the writing was the less enjoyable part, it was helpful for me to familiarize myself with the process of writing a grant proposal.
POSTERS
Below are two posters I had the pleasure to work on. The first was part of my Master's Design Project and part of a collaboration with Skeletal Dynamics, where I was working at the time. The second poster summarized the research project I worked on at the Neuroprosthetics Research Group. This poster was entered into the Research Creativity and Innovation Forum (RCIF) and won First Place in Engineering.
