aipoch/open-science

What it solves

AIPOCH Open-Science addresses the fragmentation of AI research by consolidating chats, notebooks, scripts, scientific databases, and reporting tools into a single, local-first workbench. It solves the problem of non-reproducible AI-assisted research by ensuring that execution and evidence stay together, providing traceable provenance for every generated result.

How it works

The platform uses model-agnostic AI agents that can execute Python and R code, search the web, and query scientific data sources. It operates as a local-first application where project state is stored on the user's computer. Researchers define goals in plain language, and the agents perform tasks—such as literature reviews, data analysis, and simulation—producing immutable, checksummed artifact versions. These artifacts include a "Provenance" view that exposes the producing code, execution history, and referenced inputs, allowing results to be verified or replayed in isolated environments.

Who it’s for

It is designed for scientists and researchers across various disciplines, including machine learning, statistics, life sciences, chemistry, materials science, physics, and environmental science who require reproducible and inspectable computational research.

Highlights

• Traceable Provenance: Every generated artifact is stored as an immutable version with a record of the code and evidence used to create it.
• Model-Agnostic: Supports various cloud providers (OpenAI, Anthropic, NVIDIA), custom gateways, and direct subscriptions.
• Local-First Control: Keeps application and project state on the user's machine to ensure data privacy and control.
• Integrated Scientific Stack: Combines AI agents with persistent Python and R notebooks and scientific data connectors.
• Cross-Platform: Available for macOS, Windows, and Linux.