Home

Welcome to the AI OnDemand wiki!

What is AI OnDemand?

AI OnDemand (AIoD) is a framework that enables running an expandable variety of deep learning segmentation models with ease, simplifying interacting with these models, and running them at scale on HPC. It is designed to allow the use of arbitrary user interfaces (currently with a fully-fledged Napari plugin) that plugs into a pipeline that efficiently parallelises the segmentation in almost any compute environment. More of an overview can be found in this presentation (recently given at STEP-UP 2025).

AIoD is currently comprised of 4 separate repos:

• A Napari plugin to load data, select the model, define any relevant parameters, visualise results, and ultimately submit the Nextflow pipeline...
• A Nextflow pipeline to define how to run the models, and to manage the flow of data and submitting jobs to the appropriate compute environment (HPC, cloud, or otherwise)
• A model registry (this repo!) to define what models are available, where they are located, and any parameters/config needed (which can be edited from the Napari plugin)
• A utilies repo to provide a common set of utilities for the pipeline, Napari plugin, and future interfaces to use for consistency and ease of development (primarily in I/O and preprocessing functions)

AIoD Development

AIoD was developed by the Software Engineering & AI team at the Francis Crick Institute to, first and foremost, provide a tailored tool for accelerating research. By working with various groups, primarily the Electron Microscopy STP, we have aimed to create an accessible and useful tool for users at all levels, particularly those working on the microscopes to actually generate the data.

While our approach is Crick-first, the documentation here covers what is needed for those external to the Crick to get AIoD up and running! AIoD is designed, however, to be easily extended by the community primarily through the model registry and modular nature of the pipeline. This Wiki provides information about how to do that!

Getting Started

See our Getting Started page, or checkout the sidebar if you're looking for something specific!