Hi, I'm Koral

C++ equity exchange running on AWS EC2 c6i.metal. Clients connect over TCP using FIX 4.2. Includes a price-time priority matching engine, UDP multicast market data, and SQLite event log & book persistence for crash recovery.

Analytical macro placer built for the Partcl/HRT chip placement competition. Beat the RePlAce baseline by 30% on proxy cost. Multi-phase pipeline: electrostatic GP → soft-only Adam → numerical subgradient → Lin-Kernighan k-opt → hierarchical + global LAHC.

Developed a 3D object compression algorithm using octrees, achieving up to 99% memory reduction. Built a real-time OpenGL visualizer using MVC design pattern.

A concurrent arbitrage trading engine written in Go that aggregates live order book data from multiple exchanges into a centralized Global Order Book to identify and execute profitable spread opportunities in real-time.

Built a 95% accurate neural network to classify 12 types of cancer from microarray samples. Won the Canada-Wide Science Fair and represented Canada at the EU Contest for Young Scientists (2nd place).

Collaborative whiteboard that turns voice and video into structured diagrams in real-time. Powered by Gemini 2.0 Flash and AWS Amplify Gen 2 for spatially-aware, multimodal collaboration.

Led a 5-person software team to develop a centralized government contracting platform. Built a fully responsive landing page and 10+ application pages using React with MUI and Tailwind CSS.

Built a real-time lane detection system with OpenCV canny edge detection and hough transforms. Overlayed vehicle and traffic sign tracking using YOLOv4-tiny.

Developed a medium-frequency sentiment analysis trading bot that scrapes r/worldnews. Live-tested Keras and Sklearn models reaching 12% annualized returns.

Built a hands-free typing system using MediaPipe hand tracking to control the mouse and keyboard.

A startup that transforms speech into presentation slides live using AI. Fine-tuned GPT-3.5 and used DALL-E for image generation. Deployed to prsntai.com using Vercel.

2024 UofTHacks project. Website that inputs photos and generates Spotify playlists that encapsulate the moment using GPT-4 Vision and fine-tuned GPT-3.5.

Built at DeltaHacks 2024, Bryan is an AI assistant that listens to lectures and generates real-time notes using Whisper and GPT-3.5.

Personal repo for ML/DL experiments: CPH survival models, decision trees, ResNet-50 saliency maps, and Kaggle code. Inspired by Hands-On Machine Learning.

Built an artificial neural network from scratch to classify cancer in histopathological and blood images, with a Flask web app for image-based diagnosis.

COVID-19 spread simulation using 2008 Beijing taxi data and PyGame. Users can control infection parameters and observe how the virus spreads over time.

Journey from Python to C++. Solutions and notes for major contests (AOC, CCC, USACO, CodeForces, etc.), plus scripts for stress-testing and code templates.