Bryan Zhao

Computer Scientist at the Georgia Institute of Technology

Software Engineer at the Georgia Tech Research Institute

Ex-AI Researcher at the MIT Lincoln Laboratory

Ex-Head Teaching Assistant for Machine Learning

About

Hi, I'm Bryan, a mechanical engineer turned computer scientist who is pursuing a Master's degree at Georgia Tech.

I pivoted because I wanted to design the brains rather than the brawn of systems and devices. Using machine learning and artificial intelligence, I teach robots how to learn, see, and communicate with us. I love discovering new deep learning architectures and implementing them to enhance the abilities of robots and other forms of embodied AI. Ultimately, I hope to integrate new and emerging AI technologies seamlessly into society to improve our everyday lives.

Currently, I am interning at the Georgia Tech Research Institute as a full-stack developer to strengthen my engineering skills and foundation. The applications I am building are a part of a software system that trains air force pilots using real-time simulation.

Recently, I finished working at the MIT Lincoln Laboratory where I developed AI algorithms for the Spot robot. We aimed to developed a framework that will enable agents to operate in highly dynamic outdoor environments for search-and-rescue missions by leveraging 3D scene graph technologies such as MIT Hydra.

Previously, I have revamped the vision system for a tennis-playing robot and improved its ability to detect and hit tennis balls in real-time using NVIDIA Jetson GPUs. I have also interned at BMW's IT Research Center where I conducted research on Large Language Models, enormous neural networks that are trained on vast amounts of online text data to generate and understand language. In the end, I built a prototype using GPT-3 which was OpenAI's precursor models to ChatGPT and GPT-4. Through these experiences, I aim to build a strong foundation in computer vision, natural language processing, and software systems for machine learning applications.

During my time as an engineer, I built robotic systems at Raytheon and Anduril, and I collaborated with NASA JPL to design the liquid sampler of their EELS space robot. In my free time, I like to play soccer, cook new recipes, and hit the gym. I am also rated in the 75th percentile of players on Chess.com.

Experience

Education

Georgia Institute of Technology

Master of Science in Computer Science

May 2024

Specialization in Machine Learning

GPA: 4.00/4.00

CS 7641 Machine Learning

Head Teaching Assistant | Fall 2022, Spring 2023, Fall 2023

Georgia Institute of Technology

Bachelor of Science in Mechanical Engineering

May 2022

Minor in Computer Science

GPA: 3.98/4.00

CS 4641 Machine Learning

Teaching Assistant | Spring 2022

ME 3340 Fluid Dynamics

Teaching Assistant | Fall 2020

Georgia Tech Europe

Resident Assistant | Summer 2019

Glastonbury High School

High School Diploma

May 2018

Valedictorian

GPA: 4.00/4.00

Professional

Georgia Tech Research Institute

Software Engineering Intern

Jan 2024 -
Present

  • Created a C# command line app to create, configure, and stream 200+ types of objects that each contain 100+ properties for the purpose of simulating load, unit testing, and debugging a data pipeline based on the DDS publish-subscribe protocol
  • Developing a C# GUI using Avalonia to reduce the time to create and send objects by 30% versus the command line

MIT Lincoln Laboratory

Research Intern

May 2023 -
Dec 2023

  • Implemented Python scripts to integrate OpenStreetMaps API responses into a data pipeline to generate 3D visualizations of geospatial data, convincing management of the need to extract high-fidelity maps for autonomous robot navigation
  • Created Python scripts using the Google Maps API to generate training and validation datasets containing 1000 satellite images
  • Built PyTorch computer vision models using custom datasets to extract hierarchical graphs from satellite imagery, enabling robots to use MIT Hydra to navigate and plan outdoors in search-and-rescue missions
  • Designed a Python web dashboard visualizing data with 512 dimensions from CLIP embeddings, informing a new algorithm to process robot camera views which improved task completion rates by 10% in a Habitat simulation environment
  • Developed a robot operating system (ROS) Python packages for Spot to search for objects autonomously using hierarchical graphs
  • Led a group of 5 interns to build ROS software packages, interface sensors with an NVIDIA Jetson, and 3D print parts for SpotMicro, an open-source version of the Spot robot by Boston Dynamics, for multi-robot experimentation

BMW

Machine Learning Intern

May 2022 -
Aug 2022

  • Built a Python pipeline in AWS to train OpenAI’s 175B-parameter Large Language Model (LLM) called GPT-3 that automatically classifies business documents, saving $500k and achieving 98% accuracy (+30% versus existing methods)
  • Created Python scripts with AWS Textract and AWS Translate to extract text from 500+ German PDF documents for model training
  • Developed the backend and API for Meta AI's 66B-parameter LLM called OPT to be used on-premises, enabling processing of confidential data

Raytheon

Research and Development Intern

May 2021 -
Aug 2021

  • Developed the user interface, state machine transition logic, and PID temperature control in LabVIEW for mechanically testing airplane engine parts, resulting in a system that operated continuously and autonomously for 500+ hours
  • Automated 3D modeling processes to generate CAD renders, reducing time spent on modeling by 80% from 10 to 2 weeks

Anduril

Research and Development Intern

Jan 2021 -
May 2021

  • Engineered a new testing program for the next generation of drone and launcher systems to attain a 330% increase in launch energy, creating the Solidworks design, plan for manufacture and assembly, data collection system, and testing procedures
  • Designed and managed a MS Access database to track customer requirements on a $1M+ government contract

Research

CORE Robotics Laboratory

Graduate Research Assistant

Aug 2022 -
May 2023

  • Published an IEEE conference paper with a team of 10+ students presenting development of a tennis-playing wheelchair robot used to better understand human-robot collaboration in the context of sports
  • Developed Robot Operating System (ROS) packages in C++ for a camera system to triangulate the position of a tennis ball in real-time using NVIDIA Jetsons (Linux distributed computing devices), resulting in 1.5X quality and 2X framerate of video data over existing systems to improve ball return rate to 50%
  • Processed and visualized flight data and VICON camera data in Python to create a model-based controller for a quadcopter

Georgia Tech Research Institute

Research and Development Intern

Jul 2020 -
Aug 2020

  • Contributed to a research paper analyzing bio-inspired, slip-resistant materials for gloves, robotic traction, shoe soles, etc. by generating mechanical software models in ANSYS simulating animal traction on wet surfaces
  • Researched novel computational techniques to simulate 3 key behaviors to understand water evacuation under animal paws

EPIC Laboratory

Undergraduate Research Assistant

Jan 2019 -
May 2019

  • Conducted trial investigations on a hip exoskeleton to troubleshoot stepper motor performance and perform study on timing of exoskeleton assistance during the user’s walking gait on ramps and stairs of varying steepness at varying speeds
  • Modeled exoskeleton parts to assist in the development of a 20% more compact design

Naval Submarine Medical Research Laboratory

Research and Development Intern

May 2017 -
Aug 2017

  • Prototyped a remote-controlled, Ø2ft X 6ft blimp contributing to research in a novel concept in lighter-than-air surveillance
  • Developed a new method for manufacturing an air-tight blimp envelope from thick metalized polymer film
bryanzhao@gatech.edu ⇒