Machine Learning for Poker - A complete end-to-end ML pipeline for poker game analysis and intelligent play.
The system consists of:
- Development - VS Code with Claude for application development
- Training & Serving (GCP/Databricks) - Feature generation, model training, validation with MLflow, and deployment to Unity Catalog endpoints
- ML Models - Opponent Modeling, Profit Modeling, and Policy Modeling (each with Preflop, Flop, Turn, River variants)
- Web Application - Docker container on Cloud Run with Cloud SQL, Secret Manager, and LangChain integration calling GPT-4o
PokerML/
├── README.md # This file
├── PokerML-Databricks/ # ML model training (Databricks notebooks)
│ └── notebooks/
└── PokerML-App/ # Full-stack poker application
├── engine/ # Poker game engine
├── backend/ # FastAPI REST API
├── frontend/ # React/Vite client
└── agents/ # Bot implementations
Databricks notebooks for training ML models:
- Hand strength prediction
- Opponent modeling
- Action recommendation models
Full-stack poker application featuring:
- Poker Engine - Config-driven game engine supporting Texas Hold'em, Leduc, and Kuhn poker
- Backend API - FastAPI server with ML model integration
- Frontend Client - Interactive poker table UI with ML predictions panel
- Bot Opponents - Heuristic and LLM-powered bots
# Backend
cd PokerML-App/backend
pip install -r app/requirements.txt
uvicorn app.main:app --reload --port 8000
# Frontend
cd PokerML-App/frontend/poker-client
npm install
npm run devSee the PokerML-Databricks README for instructions on running the training notebooks.
| Component | Technologies |
|---|---|
| ML Training | Databricks, PySpark, MLflow |
| Backend | Python, FastAPI, SQLAlchemy |
| Frontend | TypeScript, React, Vite |
| Deployment | Docker, Google Cloud Build |
CIS 508 Machine Learning in Business - Course Project