- Utilizing the spike tokenization algorithm from a large neural transformer model to create a transformer model for decoding neural data into finger movements for online experiments- Developed a multithreaded Python script that allowed the lab's brain-computer interface to run and switch between varied machine learning algorithms in real-time- Programmed a Python greedy channel selection algorithm that resulted in 80% power savings withnegligible performance decrease- Conducted a literature review on neural channel selection to understand and synthesize the state-of-the-art

- Created a C++ control architecture for the tethered drone's base unit, integrated sensors, and optimized power usage by adjusting parameter update speeds based on importance - Refactored the codebase to utilize a unit testing framework, catching critical bugs before field testing- Built a Simulink simulator with TCP and UDP interfaces, enabling rapid software testing and validation

- Coach students individually during office hours to identify and fill any knowledge gaps and advance their understanding of course content- Support lab sessions by aiding students in the debugging process and encouraging students to independently find novel solutions to their bugs in order to bridge the gap between theory and practice

- Learned manufacturing techniques by assembling robot components while collaborating with a small team resulting in parts completed ahead of schedule- Measured new inventory using calipers and a multimeter, ensuring parts were within allowable tolerances laid out by technical drawings, preventing costly failures of assembled products and missed deadlines- Spearheaded a new storage system that dramatically reduced the time workers spent in the warehouse