01
🌏

GenGO!

Mobile App · Research

An interactive mobile language learning app built for Android as part of a research study in the Mobile User Interfaces course at York University. GenGO! implements a spaced repetition system alongside a range of unique interaction methods — tapping, typing, voice control, and dragging — so users can study foreign languages in different ways.

The results were genuinely surprising: more interactive study styles actually yielded lower test scores, contradicting the initial hypothesis. The research findings made the project far more interesting than just shipping an app.

AndroidAndroid StudioJavaXMLUX/UI DesignUX ResearchGitHub
02
🛡️

Post Guardian

Browser Extension · Hackathon

Built for Hack the 6ix 2025, Post Guardian is a Chrome extension that helps users pause and reflect before posting on social media. It analyzes the tone of a draft in real time using the Gemini API and prompts users with thoughtful feedback — encouraging more intentional and responsible sharing online.

The goal was to blend thoughtful UX design with real-time AI analysis to promote digital mindfulness without interrupting the posting experience.

JavaScriptChrome Extension APIsGemini APIGenerative AIPrompt Engineering
03
💓

Pulsefex

Embedded Systems · Hardware

An embedded heart rate monitor built as the final project for the Embedded Systems course at York University. Pulsefex captures and displays real-time heart rate and SpO2 levels using the MAX30102 pulse oximeter and TMP102 temperature sensor, running on an STM32WB55RG microcontroller with output to a SSD1306 OLED screen.

My first embedded systems project — I was eager to get deep into the hardware side and contribute as much as possible to the circuit design and firmware.

CSTM32STM32CubeIDEEmbedded SystemsCircuit DesignMicrocontrollers
04

Baseball Oracle

ML · Full-Stack · Personal

An ML-powered prediction system for MLB games, trained on 13,000+ games (2021–2026) using a RandomForestClassifier competing against XGBoost each training run. Achieves ~60% accuracy versus the 52.5% baseline — with walk-forward cross-validation to prevent data leakage.

Includes a live FanDuel odds scraper (Playwright), a bet tracker, and weather/ballpark/umpire features pulled from the MLB Stats API and Open-Meteo. The key architectural decision: Python trains and exports the decision trees as JSON, then Node.js walks those trees directly for inference — no Python dependency at request time.

Node.jsPythonscikit-learnXGBoostExpressPlaywrightMLB Stats APIDockerRailway