● Constructed real-time ROS2 perception (LiDAR, Radar, Camera) systems for a 170 mph autonomous Indy race car in the Indy Autonomous Challenge.● Finetuned an optimized 250Hz instance segmentation neural network model and built in TensorRT C++.● Developed CUDA accelerated LiDAR/Radar-Camera projection to quickly project 3D point clouds into 2D instance segmentations for real-time 3D Object Detection.● Built LiDAR point cloud clustering node for real-time 3D Object Detection.● Created a reinforcement learning adversarial multi-agent path planner using transformer architecture for overtaking opponent vehicles in a race simulation, optimizing the autonomous Indy car's performance in dynamic, multi-agent scenarios.● Developed an OpenAI Gym environment wrapper API (Python) and ROS2 interface in Unity simulator (SVL) for training reinforcement learning path planner to race autonomous Indy car against multiple opponents.● Docker containerized opponent ROS2 systems to simulate opponent race cars within the Unity environment.● Deployed and managed reinforcement learning training on Unity simulator on Microsoft Azure Virtual Machine Servers.

● Implemented a deep learning neural network for perception (PyTorch), achieving real-time car & lane segmentation at 150hz for a 170mph autonomous IndyCar in the Indy Autonomous Challenge.● Built a scalable data labeling server using a temporal deep learning model (PyTorch) and web technologies (HTML, Flask, JavaScript) for semi-autonomous labeling of 350,000+ images with 96% accuracy.● Developed Django-Python pipeline containerized via Docker to transfer annotations between tracker & OpenCV CVAT.● Implemented a cloud dataset transfer pipeline for AWS S3 with s3fs for efficient, filesystem-like transfers of labeled datasets

● Enhanced the perception stack (ROS2, C++, OpenCV, python, TensorRT) for Ava Duo Autonomous Taxi, leveraging advanced computer vision algorithms and real-time data processing to enhance vehicle safety and navigation.● Created sensor fusion for real-time 3D obstacle detection using LiDAR, radar, camera, and deep learning, optimizing performance with CUDA and SIMD (AVX) for low latency

● Turned 10 hours of time spent trending data on a single Excel sheet into a 1-minute automated process by developing a Python dashboard to parse and process rover's diverse data logs and persist to ElasticSearch Database.● Constructed an interactive GUI using Flask, HTML, CSS, and JavaScript, enabling dynamic graph generation for rover data.● Implemented event-driven data analysis, differentiating between low-level ("Mob Events") and summarized top-level ("Drive Events") data for Mars Curiosity Rover.

● Led a team of 6 members to create a Flask/React Desktop App and meet weekly set goals within an 8-week deadline.● Designed an Electron, Python desktop app that web scrapes and compares over 100 results to find and buy the bestproduct for the user in order to automate the shopping process.● Web scraped HTML text with Selenium to find over 25 user-specified products per online store page .● Created front-end GUI with Electron.js for Shoptimum with HTML, Javascript, and CSS.● Created Flask server to connect the front and back end of the Node.js Electron app.-Taught: --Node.js --HTML, CSS --React, Electron.js --Python web scraping and text parsing --Flask app for API communication between front-end and back-end --Tensorflow Library for machine learning product recommendation with data sets

-Designed a Python Flask/React application that uses Flask to pull API data results off of Yelp and UC Berkeley dining hall API, and the React webpage recommends those chosen restaurants based on the users’ locations and food preferences by parsing the respective API pull-Used Heroku, a platform built on top of AWS, to deploy and wrap Flask/React to a single webserver/web application

-Designed a DIY wireless IR thermometer to detect temperatures for covid-19 for club members to recreate. --Thermometer designed to be controlled by and send temperature data to cell phones via Wi-Fi communication to keep both the scanner and the subject apart at safe distances(more than 6 ft apart). --The project utilizes serial communication, wifi communication, servo motor control, custom header files, and parsing data to be displayed on phone

-NASA internship to design rover using Arduino kits via California Space Grant Consortium 2020-Worked with faculty to improve rover’s mobility and regolith collection to be better previous competition’s rover designs-Lead development of all circuit and electronic payloads on the rover concept-Designed a solar charger circuit with multiple Lithium-ion cells to power rover motors and payload-Coded the rover to work with Arduino UNO Microcontroller and be controlled remotely via radio remote control.-Used H-Bridge motor controllers and coded Arduino PWM to control 12V Brushed DC Motors-Created modularly-segmented design on rover to swap the rover's battery compartment, microcontroller/processor compartment, and electronic sensor equipment when necessary -Presented Rover to NASA emploees online via ZOOM due to Covid-19