[01] About
I'm an undergrad at Georgia Tech interested in union of applied ML and robotics.
My experience revolves around 3 main parts: research lab work, club involvement, and personal projects. I've worked on data imputation ML systems for aircraft telemetry data and developing adaptive MPC systems for UAVs within research. To implementing tracking, mapping, perception, and autonomy systems into competition marine vehicles. I also spent time on personal projects exploring boundaries of NNs and robots.
[02] Currently
[03] Artifacts
A two-wheel self balancing robot. Began to get experience in owning a full stack project. Designed a cascading model based controller in simulation and implemented it in real life. Implemented a self-designed PCB, used heat set inserts to minimize vibrations, and got it all working smoothly.
ESP32 C++ LQR MPU6050 SOLIDWORKS
Designed a tracking system that sits in between filtered data and high-level semantic autonomy information to have a probabilistic approach to estimating the pose, existence, and class of objects within our USVs environment. This is a part of the software stack for the RobotX competition where we have to autonomously complete tasks such as: navigating through buoy fields filled with debris, or docking the USV in specific cleared docks in tight spaces.
ROS2 C++ Python UKF JIPDA
This project won ‘Best Hardware Hack’ at HackPrinceton F25. We developed a device to allow you to physically interact with a digital audio workstation (DAW) to make music. You can control the pitch using one hand and interact with a webcam with the other to go through a menu and tap beats into existence.
Python OpenCV MediaPipe Raspberry Pi Reaper loopMIDI OSC
Project exploring the literature on grokking through an experiment in a jupyter notebook. Grokking is when a network memorizes a small modular arithmetic dataset, sits at zero test accuracy for a long time, and then abruptly generalizes. The experiment looks inside the weights during that delay to find what is forming while the metrics still look flat, and finds the same Fourier feature circuit across all four operations.
PyTorch NumPy Fourier analysis