Building intelligent systems with deep learning, computer vision, and data science. Certified by Stanford, Harvard, and DeepLearning.AI.
I've always been drawn to strategic thinking. As a kid, I spent years competing in chess tournaments nationally and internationally, developing a mindset that thrives on pattern recognition and calculated decisions.
That mindset led me to game development — I built and released games with a friend during my indie game dev period, gathering thousands of downloads. But I wanted to go deeper: into the algorithms that make machines think.
Today, I specialize in machine learning, deep learning, and computer vision. I've completed certifications from Stanford, Harvard, and DeepLearning.AI — and I'm building projects that apply these skills to real-world problems.
A solid foundation across the ML/DS stack — from data wrangling to model deployment.
A selection of ML, data science, and full-stack projects. Filter by category to explore.
Building a real-time computer vision system that watches a chessboard through a camera, detects piece positions, and generates chess notation (PGN) live. No expensive hardware — just a camera and deep learning.
Uses Faster R-CNN for piece detection and synthetic data from Blender to train without hand-labeling thousands of images. Iterated through board warping, Hough transforms, and SuperPoint before landing on the current architecture. Streaming the build live on Twitch.
Computer vision system that watches a chessboard through a camera and generates PGN notation live. Built with Faster R-CNN and a Blender synthetic data pipeline — no hand-labeling required. Currently in progress.
Face recognition pipeline that compares a user's face against a celebrity database using CNN-based embeddings and cosine similarity to find your celebrity twin.
Standalone chess engine implementing minimax with alpha-beta pruning and iterative deepening. Evaluates board positions using material + positional heuristics.
Automated labeling tool for computer vision datasets — uses a pretrained model to suggest bounding boxes, cutting manual annotation time significantly.
From-scratch implementations of linear regression and multivariable regression using gradient descent and the normal equation. Both in JavaScript.
Full-stack group messaging app built with PHP, MySQL, and JavaScript. Supports real-time message delivery, user accounts, and persistent chat history.
Browser-based multi-person video chat using PeerJS and WebRTC. No plugins required — pure JavaScript peer-to-peer connections in the browser.
Built 4 multiplayer games — Rock Paper Scissors, Tic-Tac-Toe, Connect 4, and Chess — all in C++. Used Fast DDS (Data Distribution Service) with a publish-subscribe architecture to enable real-time multiplayer functionality across all games.
Endless driving game built in Unity during my indie game development period. Navigate an infinite road, dodge traffic, and chase your high score.
Freeroam car simulator with custom physics and 3D modeling, built in Unity. One of my first shipped games that accumulated thousands of downloads.
Infinite runner with a high-score chase. Pilot a plane through increasingly impossible obstacles. Simple, addictive, and competitive.
Web tool to look up Lenovo laptop specs and warranty status by serial number. Built with vanilla JavaScript and the Lenovo API integration.
Interactive visualization of the A* search algorithm on a grid. Lets you place obstacles and watch the algorithm find the optimal path in real time.
Formal training from top institutions in machine learning, deep learning, and AI.
Contributing to academic research in media and viewer engagement.
I'm actively seeking full-time ML Engineer and Data Scientist roles. Open to opportunities in computer vision, NLP, recommendation systems, or any data-driven team. Let's talk.