Engineering Protfolio

Purpose

I built a fully autonomous robot from scratch, integrating mechanical design, embedded systems, and sensor-based navigation into a single working system.

Description

The robot drives across a surface using a vacuum to lift sand and microplastics, then sifts the mixture to separate plastic into a collection bin while returning sand to the surface. It is built on a modified chassis with 3D-printed structural parts and controlled by an Arduino Mega managing four motors: two for drive, one for the vacuum, and one for the sifting mechanism. A BNO055 IMU and PID controller keep the robot on a straight, lawn-mower-style path.

Skills Needed/Learned

Process

I started with a problem I found interesting and worked backwards: figuring out what hardware, skills, and software I would need to solve it. With limited access to parts and tools in Bermuda, I sourced components from Amazon and repurposed household items into a functioning frame. During the school year I 3D-printed structural components requiring precise measurements. Back in Bermuda, I wired and assembled the full system, then tackled the software: coordinating four motors simultaneously, implementing a PID controller with encoder feedback, and integrating IMU-based heading correction into a unified navigation loop.

Results

The core systems performed as designed: the robot navigated in a straight line using PID control. The biggest takeaway was the experience of independently driving a complex, multi-disciplinary project from idea to working prototype under real constraints.

Purpose

I received an Arduino self-balancing motorcycle kit. Rather than using the code provided in the kit, I decided to design my own so I could learn more about control systems.

Description

This project is a desktop motorcycle running off an Arduino Nano 33 IOT with the Nano motor shield. I used Simulink and Matlab to develop a modified PID controller containing only the proportional and derivative terms. The code takes Euler angles and angular rate from the onboard IMU, runs them through unit conversions and proportional gain blocks, and converts the result to a motor command between 255 and -255. This drives a motor that spins an inertial wheel perpendicular to the motorcycle, applying an equal and opposite torque to the frame to correct for tipping.

Skills Needed/Learned

The motorcycle balancing itself by applying torque to the inertial wheel.

Purpose

I developed this fish detection algorithm during my internship at BIOS to process deep-sea baited camera footage, removing frames without movement to cut down on empty frames the human classifier has to review.

Description

This algorithm uses OpenCV to convert video into successive grayscale frames, then subtracts consecutive frames to produce a map of pixel-level change. After thresholding, blurring, and pooling, it computes the ratio of motion frames to total frames. If this ratio falls within a predetermined range (fine-tuned through testing), the pair is classified as a motion event. The algorithm then determines when each motion event occurred, applies time-based processing to filter noise, joins nearby intervals, and exports the results as a text file and motion clips via moviepy.

Left: raw footage from a camera I deployed. Right: motion map applied to the same footage.

Skills Needed/Learned

Purpose

After reading a book on the Schrodinger equation and completing a course on numerical integration, I decided to explore the connection between these two topics.

Description

This project is a Python program that uses the Verlet algorithm to numerically solve the Schrodinger equation for normalizable eigenstates. I created a position vector spanning the target interval using NumPy, then ran two initial conditions through the Verlet algorithm to find the remaining solutions. I then normalized the result and used matplotlib to plot the wave function and probability distribution as a function of position.

Skills Needed/Learned

GitHub Repository

Purpose

I designed and built an Autonomous Niskin Deployment Mechanism (ANDM) as part of my internship at BIOS. The lead researcher needed an untethered way to collect deep-sea water samples from up to 3500 meters for environmental DNA (eDNA) analysis, avoiding the cost of kilometers of cable. This eDNA data is used to determine which fish were present in a water sample within 48 hours, informing local protection plans.

Description

This design attaches a Niskin bottle to a repurposed BRUVS rig containing a float, carbon fiber rod, acoustic release, and sacrificial weight. The mechanism works by attaching the Niskin's trigger to the carbon fiber rod while allowing the Niskin itself to slide over the rod independently. While falling, the greater downward force on the center rig pins the Niskin in its open position. When the system hits the bottom, the sacrificial weight's tension slackens, reversing the force direction through the rod while the Niskin's force remains downward, triggering water capture.

Test deployment in 6 meters of water. The Niskin stays open during descent, then slams shut on contact with the bottom.

Skills Needed/Learned

Purpose

My school had an old solar car that had been sitting in a garage for 20 years. Our goal was to resurrect it while making a few upgrades.

Description

We changed the batteries and the charging architecture from parallel to series, decreasing the current running through the wire to increase safety and durability while still keeping the discharge in series. We also changed the solar and wall charge controllers to newer tech and removed many unnecessary wires. We replaced the ignition and key slot by combining them into one; finally, we replaced the speedometer and dashboard user interface.

The car being driven after restoration.

Skills Needed/Learned

Purpose

There was a persistent theft problem in my high school dorm. Milk would disappear from the kitchen in less than a day, so I decided to make a lock for it.

Description

The lid for my specific milk jug did not have any indents or a screw cap, so I could not attach one there. Instead, the lock had to be two separate parts joined together by padlocks with a bar that passed through the handle of the milk jug so it could not be lifted.

The lock in action.

Purpose

My high school did not have an EZ-bar, which I like for training triceps and biceps. Rather than buying one, I decided to make one.

Description

I bought rebar online for its low cost, grip texture, and malleability, then used a vise to bend it into equally segmented parts at a 30-degree angle. I attached barbell collars to the ends using 3D-printed adapters and metal pipe sections.