Getting Started¶

Prerequisites¶

Python 3.13+
Conda (for CUDA toolkit)
Git

Installation¶

1. Clone the Repository¶

git clone git@github.com:romain-blanchot/montecarlo-portfolio-risk-gpu.git
cd montecarlo-portfolio-risk-gpu

2. Create the Conda Environment¶

The conda environment provides CUDA toolkit (nvidia channel) and Python 3.13:

conda env create -f environment.yml
conda activate portfolio-risk-engine

3. Install Pre-commit Hooks¶

pre-commit install

Pre-commit runs ruff (lint + format), mypy, and pytest on every commit.

4. Verify Installation¶

# Run the test suite
pytest tests/

# Launch the CLI
portfolio-sim

GPU Support (Optional)¶

For GPU acceleration, install CuPy matching your CUDA version:

pip install cupy-cuda12x

Verify GPU availability:

python -c "import cupy; print(cupy.cuda.runtime.getDeviceCount(), 'GPU(s) found')"

Project Layout¶

montecarlo-portfolio-risk-gpu/
├── src/portfolio_risk_engine/    # Source code
│   ├── domain/                   # Business logic (models, ports, services)
│   ├── application/              # Use cases
│   ├── infrastructure/           # Adapters (Yahoo Finance, CPU/GPU engines)
│   └── cli.py                    # Interactive CLI
├── tests/                        # Unit + integration tests
├── notebooks/                    # Jupyter demos and benchmarks
├── scripts/                      # Benchmark scripts
├── docs/                         # MkDocs documentation source
├── pyproject.toml                # Build config (hatchling + hatch-vcs)
└── environment.yml               # Conda environment