Hi, I'm Ryan

This project leverages YOLO for detecting and tracking players, referees, and footballs in match footage, using K-means clustering to differentiate teams by t-shirt color and measure ball acquisition percentages. By tracking player movement with optical flow and perspective transformation, it accurately calculates distance covered and speed in meters.

DLW Hackathon is the biggest annual student-run hackathon in NTU. Here, a group of friends and I built a computer vision model that can identify if a hawker table is taken based on imagery. Initial simulation was built on Unity.

This is a demo of a data request access workflow. The workflow previously utilized the Databricks API to showcase the full end-to-end functionality of the workflow. However, the API has been disabled for this demo. The workflow is now a mock workflow that showcases the functionality of the workflow without the need for the Databricks API.

Based on agent-based HDL generation via LLMs, D3code is a finetuned Llama 3.1 (8B) model that utilizes GRPO techniques to create a reasoning model that is capable of generating D3.js code snippets based on user data input. It is designed to assist users in generating visualizations quickly and efficiently.

SweatPals is a social workout app designed for fitness enthusiasts. It allows users to connect with likeminded individuals and find nearby facilites. The app features a user-friendly interface, real-time notifications, and a robust backend to handle user data securely. Utilized government APIs for fitness facilites.