Projects
StratBench
Active DevelopmentProfessional-grade algorithmic trading backtesting platform with AI-powered strategy generation, ML model training, and advanced technical analysis
StratBench enables users to build, test, and analyze trading strategies on historical market data. The platform combines AI-powered natural language strategy generation, machine learning model training (9 production-ready models), and advanced technical analysis in a modern web application. Features include interactive stock charts with technical indicators, a strategy builder that converts natural language descriptions into executable trading rules, a comprehensive backtesting engine for both rule-based and ML-based strategies, and a stock screener for discovering opportunities.
🎯 Core Features
• AI-powered strategy generation
• 9 ML models (RF, XGBoost, LightGBM, etc.)
• Technical indicators (SMA, EMA, RSI, MACD, Bollinger)
• Stock screener & multi-ticker comparison
⚙️ Tech Stack
• Frontend: Vite + React 18 + TypeScript + Tailwind
• Backend: FastAPI (Python 3.12)
• Database: Supabase PostgreSQL
• Data: Polygon/Massive API + S3 Parquet
📊 Current Status
• 7 tickers loaded (AAPL, AMZN, GOOGL, META, MSFT, NVDA, TSLA)
• 3 months historical data (Oct-Dec 2024)
• 228K+ minute-level bars
• Testing stage, ready for expansion
🚀 Architecture
• Frontend-first design (4/5 pages work offline)
• Hybrid data access (Supabase + backend API)
• Minute-level precision backtesting
• Production-ready ML models
Past Projects
Machine Learning for Trading
A collection of machine learning implementations and algorithmic trading projects. Includes custom learner implementations (Decision Trees, Random Trees, Bag Learners), Q-Learning for trading strategies, and market simulation tools built in Python.