v0.1.5 — Production Ready

Systematic Trading
Democratized.

The library I use to research, backtest, and run systematic strategies.Pure Python, explicit assumptions, no black boxes.

MIT License
Python 3.10+
100% Vectorized
pyfundlib — python
# Install from PyPI
pip install pyfundlib

# Or install with poetry
poetry add pyfundlib

What pyfundlib ships with

Covers the path from a quick notebook idea to a production deployment.

Backtesting & Research

  • Vectorized engine (100× faster)
  • Walk-forward analysis
  • Deflated Sharpe & PSR
  • Monte Carlo stress testing

Portfolio Construction

  • Risk Parity (ERC, HRP)
  • Black-Litterman
  • Mean-CVaR optimization
  • Hierarchical Risk Parity

Execution & Live Trading

  • Alpaca & Interactive Brokers
  • Paper + live modes
  • Slippage modeling
  • Multi-asset support

Machine Learning

  • MLflow model registry
  • LSTM, XGBoost integration
  • Rolling feature pipeline
  • Reproducible workflows

Where it actually fits

A few realistic places pyfundlib tends to land in people's workflows.

Research notebook to production

Start in a Jupyter notebook, end with a deployable, versioned strategy that survives code review.

Signal research and factor testing

Quickly try new signals on clean data without rewriting backtest infrastructure for every idea.

Existing stack integration

Drop pyfundlib into an existing Python stack instead of replacing everything with a framework.

What it is actually good at

A few honest highlights from how PyFund is meant to be used day to day.

Research without rebuilds

Iterate on signals and portfolio logic without constantly rewriting backtest infrastructure.

Production-minded from day one

The same code paths you use in notebooks are designed to end up in CI and production.

Transparent, debuggable math

Everything is plain Python and NumPy. When something is off, you can actually inspect it.

Our Philosophy

No Jupyter Lock-in

Works in scripts, Docker, cloud functions, and production servers. Not just notebooks.

Transparent

Every calculation is vectorized, readable, and testable. No compiled C++ black boxes you can't inspect.

Zero Lock-in

MIT licensed. Fork it. Deploy it. Own your IP. No subscriptions.