User:Bryan Heddle

Bryan Heddle is currently a third year student at the University of Western Ontario, pursuing a bachelors degree in mechatronics and artificial intelligence systems engineering. He is passionate about robotics, embedded systems, software development, artificial intelligence, and just about anything regarding engineering! On his free time he enjoys spending time with his dogs and family.

Mechatronics Systems Engineering + Artificial Intelligence Systems Engineering

• 2022 - 2027
• First Year average: 81%
• Second Year average: 86%
• Deans Honor List (2022-2024)

• Manage testing and QA for key LLM tasks including Retrieval-Augmented Generation, bug fixing, unit testing, and solution generation, improving overall code quality and performance.
• Conduct edge-case testing and performance analysis, identifying key improvements that increase model accuracy and enhance complex query handling.
• Ensure continuous model refinement and stability by tracking LLM tool logs to identify errors and performance bottlenecks.

• Taught children aged 5-14 about all things STEM related through small projects, whether it was circuits for turning on LED's to code block programming.

• Initiated development of a real-time video streaming app using WebRTC and GStreamer in Python, which delivered high-quality streams to a React front-end.

• Programmed an IMU, GPS, and their data pipelines, also developed detailed system monitoring.
• Led optimization of data pipelines using CANbus communication, ensuring noise-free data flow.
• Increased ease-of-use by integrating ROS2 for data management from multiple sensors and controls
• Integration of LiDAR, IMU, and camera systems utilizing Jetson Orin Nano to create simultaneous localization and mapping program.
• Worked with many microprocessors extensively, including but not limited to esp32's

• Designed antenna mount for rover in Solidworks

• Currently developing rocker bogie differential for rover

• Developed a MongoDB database for storing sensor data and integrated a Python back end and a React front end for sending and displaying the data, also utilizing Docker.
• Designed a custom PCB for integrating the aircraft’s sensors, enabling data receiving and connection to the onboard control system, improving system reliability and reducing wiring complexity.

• Develop commands and data handling architecture for the main computer using C++
• Communicate with electrical sub-team to plan out power supply for main computer
• Maintain subsystem documentation

• Provided safe accompaniment for individuals who felt unsafe walking home at night.
• Conducted regular campus patrols during scheduled shifts to promote a secure and welcoming environment.

• Competed in re-engineering category at the Western Engineering Competition and won first place [2023].
• Competed at Ontario Engineering Competition, re-designed a ferry to hold more passengers and ensure safety [2023].
• Competed in the senior design category at Western Engineering Competition [2024].

To view technical project information, check out my Github repositories!

• Modified open source robotic arm design using SolidWorks, optimizing for 3D printing and assembly.
• Developed C++ servo scripts with Arduino IDE to parse incoming data and control the motors of the robotic arm
• Designed a C# Windows application to facilitate data transmission to an Arduino UNO via Bluetooth, prioritizing efficiency and practicality
• Configured an electrical circuit connecting an Arduino UNO with servo and stepper motor drivers, and a Bluetooth chip.

• Implemented computer vision model training methods based off YOLOv3’s research paper.
• Learned YOLOv3’s darknet architecture for training weights on a pre-made dataset
• Optimized hyperparameters based on techniques from YOLOv3 research paper. Maintained 51.2% mean average precision

• Developed multiple machine learning models for fraud detection, including Linear Regression, Logistic Regression, Random Forests, AdaBoost, XGBoost, and Neural Networks using scikit-learn and PyTorch.
• Implemented multi-layer neural networks with PyTorch, on complex fraud detection tasks.
• Optimized hyperparameters of multi-layer neural network with cross-validation, improving model accuracy to 99% and precision to 83%.
• Preprocessed large datasets with Pandas and NumPy, ensuring clean, feature-engineered input for models.

Languages: Python, Java, C++, HTML/CSS, React, JavaScript, SQL, Bash, XML

Technologies/Frameworks: VS Code, Amazon Web Services, Docker, Flask, NodeJs, Linux, GitHub, APIs, SCIKit, Tensorflow, PyTorch, Keras, Panda, Numpy, Hugging Face, OpenCV, ROS2, Excel, Word, Powerpoint

Electrical: Altium, Eagle, Soldering, PC Design and Assembly, PCB Testing and Debugging, Function Generators

Mechanical: Solidworks, Structural Design and Analysis