Skip to main content

For AI agents: the complete documentation index is at llms.txt. Markdown versions of pages are available by appending .md to the URL or sending Accept: text/markdown.

Financial Machine Learning

Orion is built on the principle that Financial Machine Learning can make portfolio management more accessible onchain. By combining data-driven insights with automated execution, Orion supports investment strategies that a broader range of participants can access.

Why It Matters

In traditional finance, advanced portfolio optimization tools are commonly reserved for institutions and hedge funds.

Orion brings these capabilities onchain, so:

  • Everyday users can access strategies that simplify allocation decisions.
  • Experienced managers can enhance performance with data-driven insights.
  • All participants can benefit from automated allocation.

Two Core Approaches

Orion supports two main styles of portfolio management:

  1. Passive Portfolios - Simple, rules-based strategies (e.g., indexes) that automatically rebalance.

    • Suitable for users who prefer hands-off, predictable exposure.
  2. Active, Machine-Learning-Driven Strategies - Models that adjust allocations in response to data.

    • Suitable for users seeking more dynamic, performance-oriented portfolios.

Powered by skfolio

Under the hood, Orion integrates skfolio, an open-source portfolio management library for quantitative finance. This allows vault strategies to leverage advanced statistical techniques, from covariance analysis to risk-adjusted optimization, while keeping execution verifiable and onchain.

What This Means for Users

With Financial Machine Learning-powered vaults, a user can choose a strategy that matches their preferences:

  • Stay passive with transparent, rules-based allocations.
  • Use active, private models that seek to improve risk-adjusted returns.

In both cases, Orion handles execution in the background, rebalancing at set intervals, keeping costs lower through batching, and protecting strategy details where configured.