Skip to content

Introducing Project Fleming: Enhancing Code Reuse and Efficiency with AI Discovery

GlassImage

Today, Shell’s Inner Source Enablement Team has released key packages for a new AI-powered Discovery Tool. These packages are available for Open Source use.

The Discovery Tool was created to address a specific Inner Source challenge: while setting a code repository to “open” is a good start, integrating it into a workflow using natural language search takes it further. Discovery is not just about finding what you expect; it’s also about uncovering what you may not have anticipated. By leveraging AI, we make this possible. The tool can search for code as well as Stack Overflow entries. Our goal is to ensure that no developer leaves empty-handed, ultimately enhancing efficiency. Already, the tool has helped developers identify opportunities for code reuse.

The idea for this tool originated from PhD research at Loughborough University, which demonstrated how powerful AI could be in classifying information, even when it was in an unstructured state.

This tool would not have come to life without the skill, dedication, and creativity of the Inner Source Enablement Team and all those who supported us. Innovation rarely happens in isolation, and this was no exception. The team not only brought the concept to life but made it efficient, for example, by running it on a CPU instead of a GPU setup. Furthermore, we have striven to make it model-agnostic, thus allowing users to integrate the LLM of their choice.

Specific thanks go to the following individuals:

Shell

  • Benedict Butcher
  • Amber Rigg
  • Tugce Ozberk Yener
  • Oliver Carr
  • Gabriel Molina
  • Tom Lewis
  • Anusha Modwal
  • Michelle Bhaskaran
  • Herman Kruis
  • Niranjan Girhe
  • Neethi Mary Regi
  • Namitha Raveendranathan.

Loughborough University

  • Professor Jenny Harding
  • Dr Diana Segura-Velandia

We also could not have achieved this without management support from Bryce Bartmann, Dan Jeavons, Warren Harding, Adam Jordan and Karina Fernandez.

We encourage all colleagues and potential collaborators interested in Inner Source and Open Source to explore the tool. We welcome your contributions as we continue its development. We also look forward to unexpected insights and new features that others may add.

As always, we value your feedback - feedback is a gift.