Leading a team of 6 developing a robotic solution to California's exploding population of purple sea urchins. On the technical side, I spearhead the software development for the project, focusing on developing a computer vision system to run in real time to detect and track purple sea urchins using the camera onboard the robot.

Spring 2023 - Computer Vision (CSCI153)Fall 2022 - Computability and Logic (CSCI081)Fall 2021 and Spring 2022 - Principles of Computer Science (CSCI060)Spring 2021 - Introduction to Computer Science (CSCI005)

Action Attention Project -- Fall 2022 to Spring 2024Computer vision research project aiming to quantify how action recognition neural networks make action classification decisions. Working with multiple action recognition neural networks, production scale datasets, and generated adversarial data to quantify the static and dynamic biases of these models.SMILER Project -- Spring 2022I worked on the Saliency Model Implementation Library for Experimental Research (SMILER) project in the Lab for Cognition and Attention over Time and Space (Lab for CATS) at HMC. I adapted the previously Linux-only tool to run Python and MATLAB saliency models in Windows/WSL.

Working with collision risk team to predict future actions of other road users

Worked with the Simulation Acceleration and Services team on a high priority simulation accuracy initiative. Created a spoofer node using C++ and ROS to recreate messages published by the AV's Perception stack in simulation, making the simulator a more accurate and better testing environment for other teams in Prediction and Planning. Validated spoofer accuracy with a Python pipeline to compare spoofed messages to those created by the AV itself. Gave presentation to 80+ attendees summarizing my intern project and its impact on Cruise.

I led a 6 person Computer Vision subteam which adapted an open source image labeling software for MuddSub's use. We distributed images to remote labelers through a web server, simplified the labeling process, and verified accuracy of image labels. Additionally, I researched range estimation methods for our underwater robot.

Working full time over the summer, I implemented and tested Python and C++ code using ROS for robot controls and sensors including a DVL, an IMU, and a depth sensor. I also developed simulated ”perfect nodes” of robot software subsystems to provide a testing environment for experimental versions of subsystems.