Brawl Stars is fun but I got tired of endlessly guessing which brawlers would work best on specific maps, so I built a tool to do the heavy lifting for me. It grabs the data, crunches numbers, and shows you a nice visual with the best choices to maximize your chances of winning on any map. No more guesswork! Now with over 20k users and growing!

A browser extension I created that helps save web content in clean, formatted Markdown. Built on Mozilla's Readability.js with custom enhancements for developers like improved table formatting, code block language detection, and batch URL processing. Perfect for saving documentation, research materials, and technical content for offline reference or LLM assistance. Hit 2500+ active users in less than a month! Currently at 30,000+ users!

A badminton analytics platform I built using Ultralytics YOLOv8 to detect players and shuttlecocks from match footage, plus a CNN-based court keypoint model for court mapping. I then quantify the video with geometric and kinematic calculations, using player-height-based distance estimation to derive metrics like shuttle speed, positioning patterns, and other tactical insights.

A Chrome extension I built to convert GitHub issues, pull requests, and discussions into clean, copyable Markdown with one click. It preserves rich formatting like code blocks, links, images, and nested replies, and adds configurable export controls for timestamps, bot comments, event verbosity, frontmatter copy, keyboard shortcuts, and .md downloads.

A VS Code extension I created to convert workspace files into well-structured prompts for LLMs. It features fuzzy search, smart filtering with .gitignore support, and Claude-optimized XML output formatting. Perfect for developers working with AI models who need to quickly reference multiple files in their prompts without manual formatting. Already has 600+ installs and growing!

A computer vision tool I built to automate brawler detection from SpenLC's draft guide images. It uses OpenCV template matching with multi-scale detection, IoU-based NMS, and ROI-specific thresholding to extract structured data from 24+ guide images. The JSON output feeds directly into Brawlytics, enabling the 'SpenLC' filter to layer curated pro picks into the ranking algorithm.

Tap, speak, go: a hackathon-built OpenAI assistant that captures fleeting thoughts, auto-sorts them into lists, and generates one-tap actions like purchase links. Built during the 757 Build Weekend Hackathon using Next.JS, TypeScript, and OpenAI API.

A personalized anime and manga recommendation platform I built to help users discover content tailored to their unique tastes. It connects with AniList to analyze your ratings and preferences, then suggests new shows and manga you're likely to enjoy.

One of the tools I made while writing code using Claude.AI. I found that I was constantly having to paste chunks of my code into Claude so that it wouldn't turn into an attachment, so I came up with an easy to use Google Chrome Extension, Claude Paste, to solve that problem.

This website itself is a side project! It took a while to create using React + Next.JS (with TypeScript) + Framer Motion and Tailwind CSS. I try to update it as often as I can.