OptiQuant is an end-to-end quantitative trading intelligence platform — built to do what hedge funds do: engineer signals, blend models, eliminate lookahead bias, and surface alpha.
Most stock prediction tools are toy projects. OptiQuant is not.
It's a full quantitative research pipeline — from raw OHLCV data → feature engineering → ensemble signal generation → backtesting with institutional-grade metrics. Deployed on AWS EC2. Containerized with Docker. Built for real analysis, not vibes.
┌──────────────────────────────────────────┐
│ RAW OHLCV DATA │
│ (CSV Upload or Live Manual Input) │
└─────────────────┬────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ DataPreprocessing.py │
│ ┌────────────────────────────────────┐ │
│ │ Feature Engineering (12+ signals) │ │
│ │ • Momentum (1d, 5d, 20d returns) │ │
│ │ • Volatility (rolling std) │ │
│ │ • RSI (14-period) │ │
│ │ • Risk-Adjusted Metrics │ │
│ │ • Lookahead-BiasFree Construction │ │
│ └────────────────────────────────────┘ │
└─────────────────┬────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ ENSEMBLE MODEL LAYER │
│ │
│ LightGBM ──┐ │
│ CatBoost ──┼──► Weighted Blend ──► │
│ RandomForest┘ Signal Score │
└─────────────────┬────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ SIGNAL REFINEMENT │
│ • Volatility normalization │
│ • Smoothing function │
│ • Daily cross-sectional ranking │
└─────────────────┬────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ STREAMLIT DASHBOARD │
│ • Backtesting engine + metrics │
│ • Live prediction mode │
│ • Interactive performance charts │
└──────────────────────────────────────────┘
Three powerful models blend their predictions via weighted averaging — not just picking one winner. This reduces variance and improves signal stability across different market regimes.
| Model | Role |
|---|---|
| LightGBM | Fast gradient boosting; excels at large-scale tabular features |
| CatBoost | Handles categorical splits natively; robust to outliers |
| Random Forest | Decorrelated trees; strong regularization via bagging |
All features are constructed using only past data, no future values leak into predictions. This is one of the most common and fatal mistakes in quantitative research. OptiQuant handles it correctly.
Engineered signals include:
- Rolling momentum (1d, 5d, 20d returns)
- Historical volatility (rolling standard deviation)
- Relative Strength Index (14-period RSI)
- Risk-adjusted return metrics
- Volume-weighted signals
Upload any CSV with historical stock data and instantly evaluate the strategy's historical performance.
Input a single stock's current market data and receive a real-time signal score — ranked against the model's expected universe.
| Metric | What It Measures | Why It Matters |
|---|---|---|
| Sharpe Ratio | Risk-adjusted excess return | The gold standard — penalizes volatility |
| Calmar Ratio | Return ÷ Max Drawdown | Performance during the worst periods |
| CAGR | Compound Annual Growth Rate | Annualized wealth accumulation rate |
| Max Drawdown | Largest peak-to-trough loss | Worst-case scenario exposure |
| Win Rate | % of profitable days | Consistency of positive returns |
Main Dashboard
Performance Metrics & Cumulative Return vs. Benchmark
| Layer | Technology |
|---|---|
| ML Models | LightGBM, CatBoost, Scikit-learn (RandomForest) |
| Data Processing | Pandas, NumPy |
| Frontend | Streamlit |
| Deployment | Docker + AWS EC2 |
| CI/CD | GitHub Actions |
| Serialization | Joblib (.joblib model artifacts) |
# Clone the repo
git clone https://github.com/RudraDudhat2509/OptiQuant.git
cd OptiQuant
# Install dependencies
pip install -r requirements.txt
# Launch the app
streamlit run app.py# Build the image
docker build -t optiquant .
# Run the container
docker run -p 8501:8501 optiquantThen open http://localhost:8501 in your browser.
OptiQuant/
│
├── app.py # Streamlit frontend — all UI logic
├── DataPreprocessing.py # Feature engineering pipeline
│
├── model_lgbm.joblib # Trained LightGBM model artifact
├── model_cat.joblib # Trained CatBoost model artifact
├── model_rf.joblib # Trained RandomForest model artifact
│
├── notebooks/ # Research & training notebooks
├── data/ # Sample datasets
│
├── Dockerfile # Container definition
├── requirements.txt # Python dependencies
├── .github/workflows/ # CI/CD pipeline (GitHub Actions)
└── .gitignore
Your uploaded CSV must contain the following columns with at least 60 trading days per stock for reliable signal generation:
date, open, high, low, close, volume, Name
| Column | Type | Description |
|---|---|---|
date |
YYYY-MM-DD |
Trading date |
open |
float | Opening price |
high |
float | Daily high |
low |
float | Daily low |
close |
float | Closing price |
volume |
int | Shares traded |
Name |
string | Ticker or stock name |
This tool is built for educational and research purposes only. It does not constitute financial advice. Past backtested performance does not guarantee future results. Always do your own due diligence before making investment decisions.
- Add SHAP explainability — show which features drove each prediction
- Support real-time data ingestion via Yahoo Finance / Alpha Vantage API
- Portfolio-level backtesting with position sizing and rebalancing
- Sector-neutral signal construction to remove market-wide bias
- Export detailed backtest reports as PDF
Built with Python, gradient boosting, and a healthy obsession with avoiding lookahead bias.