Home About Projects Contact

Things I've Built

ML models, AI bots, data pipelines, games, and tools — a mix of research-grade work and side projects I couldn't stop building.

🎵
Python ML Spotify API SQL

Million Spotify Playlists

Led a student team to predict playlist success on a 36 GB dataset of 20M songs from 2M artists. Engineered a sequential cache for 10× speedup in random data access. Applied EDA, Clustering, Logistic Regression, CART, LDA, Random Forests, and Ridge Regression — achieving 89% average accuracy.

🌍
PyTorch CNN Google Cloud

GeoGuessr AI

Convolutional Neural Network that identifies GPS coordinates from Google Street View images. Collected 50K images via custom data mining techniques and trained the CNN to 85% accuracy on large coordinate bins.

🚀
PyTorch GAIL RL

AbiGAIL

Inverse Reinforcement Learning bot for Rocket League using Generative Adversarial Imitation Learning (GAIL). Trained on thousands of high-level and professional matches to imitate expert behavior via the RLBot API.

🤖
Python ChatGPT DALL-E 3 Discord API

Gnomebot

Social Discord bot maintained for 4+ years. Uses GPT and DALL-E 3 for NLP conversations and AI image generation. Creates quote-based inspirational images, tracks air quality, integrates with Google Docs/Sheets APIs. Self-updating scripts keep it continuously live.

🎧
Java OpenAI API Android Spotify API

AI Playlist Generator

Android app + Google Assistant integration that generates a personalized Spotify playlist from a voice command. Uses ChatGPT to select songs matching your taste and a description ("Get music for walking on the beach"), then populates the queue via the Spotify Web API.

🧠
PyTorch Transformers GAIL / VAIL

Transformers for IRL

Research project using GAIL and VAIL algorithms with Transformer-based reward networks for Inverse Reinforcement Learning. Tested on the LunarLander-v4 AI Gym environment. Transformers produced more stable training, though converged more slowly than linear networks.

🎲
Python TCP Sockets PyGame Android

Codenames Online

Full implementation of the Codenames board game using TCP Sockets and Pygame — plays exactly like the original. Comes with an Android companion app for local multiplayer, letting players use their phones as the board display.

🟦
Python NEAT RL Numpy

Tetris Reinforcement Learning

Built a Tetris engine from scratch and implemented several genetic algorithms by hand to train an agent to play. Used the NEAT-Python library for the NEAT genetic algorithm. Models learned to survive for several lines but lacked long-term strategy.

🐉
Unity AR C#

D&D AR Battle Map

Augmented Reality app that displays and moves player character tokens on a real-world surface, replacing physical miniatures in Dungeons & Dragons. Players can scale and reposition characters for immersive tabletop play.

✈️
Python SIMIO Simulation

SFO Passenger Traffic Sim

Discrete-event simulation of international terminal passenger flow at SFO using real TSA and flight arrival data. Modeled de-boarding distributions, gate assignments, citizenship status, and time-of-year factors to estimate international passenger wait times.

Have a project in mind? Let's talk.

I'm always looking for interesting problems — especially in ML, data, and systems.

Contact Me Learn More About Me