ado: a Python framework for computational experimentation and benchmarking

The Accelerated Discovery Orchestrator (ado) is a Python package that addresses a recurring challenge in research software development: implementing common capabilities for design of experiments (DoE) and execution of related computational experiment campaigns. These cross-cutting capabilities span methodology (design-space specification, sampling, analysis), interface (CLI and configuration management), execution (parallel and scale-out), and data (sharing, provenance, and reuse). ado delivers these capabilities across domains through a lightweight plugin model, where integrating new components can be as simple as decorating a Python function. This is enabled by ado’s core abstraction: the Discovery Space. Out-of-the-box, ado includes state-of-the-art optimization algorithms and predictive modeling tools, alongside experiments targeting foundation-model performance. Our aim is for ado to become a focal point for developing and consuming advanced capabilities for defining and executing experiment campaigns.