Open-source quant tools surfaced

Quantitative trading code that once sat inside firms is now showing up as open-source software, with one new Python package, AutoHedge, offering a multi-agent trading stack through GitHub and the Python Package Index. (github.com, pypi.org) AutoHedge says it splits the job like a trading desk: a Director agent generates ideas, a Quant agent analyzes them, a Risk Management agent sizes positions, and an Execution agent places trades. The GitHub repository showed about 1,200 stars and 219 forks when checked on April 14, 2026, and the Python package page lists version 0.1.6, released on February 17, 2026. (github.com, pypi.org) The project’s own documentation says current live support is on Solana, with Coinbase marked “coming soon,” and installation starts with `pip install -U autohedge`. Its package page also lists environment variables for Jupiter and OpenAI or Anthropic keys, plus a wallet private key for trading. (pypi.org, github.com) In quantitative finance, “backtesting” means replaying old market data to see how a strategy would have behaved before risking real money. “Live trading” means sending orders into actual markets, where timing, fees and exchange rules can break systems that looked fine in simulation. (nautilustrader.io, docs.rs) That gap is where Nautilus Trader is positioning itself. The project says it uses one event-driven engine for both simulation and live deployment, with Rust handling the core engine and Python used for strategy logic and orchestration. (nautilustrader.io, pypi.org) Nautilus Trader’s site says the engine supports multi-asset, multi-venue trading, nanosecond-resolution simulation and advanced order instructions such as one-cancels-the-other and one-triggers-the-other. On April 14, 2026, its homepage also showed more than 21,000 GitHub stars, about 979,000 downloads and roughly 5,000 Discord members. (nautilustrader.io, github.com) A third item circulating with the code links was a finance machine-learning slide deck hosted as a free PDF. The document, titled “Artificial Intelligence and Machine Learning in Finance,” is credited on its cover to Mirko Polato and dated December 16, 2020. (makgyver.github.io) The PDF is not trading software. It is a teaching document that walks through machine-learning use cases in finance, including fraud detection, robo-advisory, algorithmic trading and reinforcement learning, which it describes as learning actions from rewards instead of labeled answers. (makgyver.github.io) Taken together, the links point to three different layers of the same workflow: an agent-based strategy wrapper, a production trading engine and a study guide for the methods behind them. They also show how much of the quant-tool chain is now being packaged for download instead of kept behind a firm’s firewall. (github.com, nautilustrader.io, makgyver.github.io)