Projects

Multi-Agent RL Trading System

Reinforcement learning agents that learn to trade in simulated stock markets. Uses PPO to train multiple agents competing against each other, with realistic price impact modeling and portfolio management.

Results:

  • Agents learn stable trading strategies with consistent returns
  • Shows realistic market impact understanding through price correlations
  • Adapts behavior in real-time to market volatility
Reinforcement Learning PPO Algorithm Multi-Agent Systems OpenAI Gym
View Project

Interactive LLM Explainability Dashboard

Interactive dashboard for exploring and visualizing how language models work internally. Makes complex neural network behaviors interpretable through attention heatmaps and text embedding visualizations.

Features:

  • Text embeddings visualization using SentenceTransformer with t-SNE
  • BERT attention heatmaps showing token-to-token attention patterns
  • Interactive interface for exploring model internals
Interpretability Attention Visualization BERT t-SNE
View Project