Built "GpuCompute," a GPU kernel performance analysis plugin for Google's Perfetto, now merged upstream into the open-source project. Designed to feel familiar to developers coming from NVIDIA Nsight Compute, it combines a cycle-accurate kernel timeline with structured per-kernel metrics (speed-of-light throughput, launch statistics, occupancy, workload analysis) in one view, with baseline support to diff relative and absolute metrics across kernel runs. Selecting a kernel on the timeline automatically opens the Compute tab, and a sortable results table makes it easy to spot the best- and worst-performing kernels across a whole workload. Built to support multiple GPU vendors/architectures through modular metric registries and terminology mappings, and later extended with automated bottleneck detection for CUDA kernels while at Rivos (acquired by Meta).

DemoCode

Completing the MIT AI course provided me with a solid foundation in machine learning principles and practical applications. Through projects like building an Automatic Review Analyzer, Digit Recognition using Neural Networks, and applying Reinforcement Learning, I gained hands-on experience in implementing models such as linear classifiers, neural networks, and clustering algorithms. The course sharpened my skills in Python, TensorFlow, and model evaluation, deepening my understanding of key concepts like gradient descent, regularization, and feature engineering, and preparing me to tackle real-world AI challenges.

DemoCode

Completing my first year of engineering with the UVic Formula SAE team was a transformative experience. Despite a last-minute battery issue that forced us to withdraw from the Michigan competition, my work on the controls system team provided hands-on experience in low-level C programming, CAN networks, and data analysis to operate and safely control a high speed EV. This opportunity elevated my skills to industrial standards, deepening my understanding of practical engineering solutions in the high-stakes environment of motorsports.

DemoCode

This project aimed to create an independent AI agent capable of playing the Flappy Bird game. Leveraging reinforcement learning techniques, the AI employs a neural network to interpret real-time game states, transforming them into strategic actions. With each generation, the dataset rapidly refines, empowering the AI to consistently score high. Notably, the video showcases the evolution from a challenging initial dataset to an AI that achieves enduring success within just few generations.

DemoCode

Project features include multidirectional swerve drive, adaptable linear extension, efficient intake, automated game piece alignment, dynamic wrist control, and precise outfeed pacing. Programming highlights involve field-oriented swerve drive, auto-balancing endgame code, PID-tuned motor control, and sensor-driven automation. Streamlined autonomous routines interpret JSON data for seamless movement and control.

DemoCode

A self driving car game that allows you to race against your very own neural network that you can develop ove time. The game uses advanced physics for movement and virtual distance sensors for providing a range of distances and feedback for the AI car being developed. The site even lets you save previous networks and delete them from your local storage. It even includes a network visualizer that can show you what the car you are up against is processing based on weights and biases.

DemoCode

Welcome to the centralized Media tech Application, where you can showcase all your projects from application and programming to robotics and electronics. First Journey allows you to create an account, make posts, edit and delete them, as well as like and comment. Some neat features include searching for specific posts by Title and or Tags. Make sure to sign up and discover some other cool features.

DemoCode

As lead programmer on FRC team 6390 Hephaestus, I was responsible for programming our robot named Kevin. Some key functionalities include vision tracking, autonomous routines, and robot subsystems such as shooting, climbing, intaking and much more. This was all done by programming a sequence of motors, servos, pneumatics, and cameras using Java as well as creating a remote controlled operating interface with a series of different controllers from Joysticks to Xbox controllers.

DemoCode

I'm proud to say I have programmed the first Swerve Drive robot in British Columbia, Canada. Programming it was extremely difficult but it has its advantages such as moving in any direction as well as rotating as it drives. This allows the robot to have a lot of movement and control and was programmed in a field oriented way since each wheel or module can move back and forward and spin clockwise and counter cloockwise infinitely.

DemoCode

While working at Stylish Tile as a Web Developer I was responsible for creating and developing a responsive and interactive website from scratch. This includes creating image libraries, and animations.

DemoCode

Minlexa was my project submission for the Replit Summer Amazon Alexa Hackathon. It allowed you to ask alexa any question related to crafting in Minecraft and would inform you about all the necessary materials needed to craft the item, as well as give you a brief description about it, and a visual display.

DemoCode

Developed using the A* and Dijkstra algorithms to find the shortest path possible to a desired setpoint. It was developed using Java Swing with multiple GUI features. It was even expanded upon into a full scale application to develop robot coordinates which can be used to automatically create an autonomous routine which can be seen in the video demonstration.

DemoCode

Using Machine Learning along side Python and OpenCV I was able to develop a face recognition system that would auto detect anyone it can recognise from the pictures uploaded to the program trainer. From there I further developed it during the pandemic to recognise whether or not someone is wearing a mask which can be prove useful for enforcing specific rules and laws.

DemoCode

Exhaust was my Project Tech Conference Hackathon submission. It is a carbon emission tracking application which allows you to visualize all your data with graphs and charts as well as global statistics showing you all emissions from other participating countries. With features such as google maps active tracking using the maps api for more accurate results.

DemoCode