ML Playground

A machine learning playground for experimenting with various ML models and techniques.

🎮 Web Demos

Open any HTML file in visualizations/ in your browser.

Q-Learning Pac-Man Demo

Interactive Q-Learning reinforcement learning demonstration featuring a Pac-Man agent learning to navigate a maze and collect dots while avoiding ghosts. Built with HTML5 Canvas and JavaScript, this demo showcases RL concepts including Q-value visualization, hyperparameter tuning, and training progress monitoring.

Modeling Projects

Generation Models

Image generation and reconstruction models including autoencoders, VAEs, and planned implementations of GANs, Stable Diffusion, and Diffusion Transformers (DiT).

LLM Embedding

Sentence embedding learning using SimCSE approach.

LLM Post Training

Docker-based environment for LLM post-training experiments.