IBM Research Molecular Discovery Tool • Unleashing the Power of Generative AI for Molecular Lab Scientists

In the early months of 2022, IBM Research had a valuable asset at its disposal: a selection of proprietary AI models that could accelerate molecular discovery. These models were technologically advanced, but only interactable in code form—which wasn't particularly user-friendly for lab scientists who lacked coding knowledge. Seeking to bridge this gap, IBM Research engaged us to create an intuitive interface that would enable these scientists to leverage the AI models effectively, thereby facilitating the creation of new molecules that could find drugs for diseases, sustainable materials and much more. The objective was not just to make the models accessible, but to ensure their seamless integration into the existing scientific process.

To kick off the experiential strategy, we began an in-depth exploration of the scientists’ culture through interviews and workshops. During this process, we identified three critical principles that would serve as the foundation of our approach: flexibility, play and collaboration. We recognized that each scientist had a unique problem-solving approach, appreciated the importance of fun and exploration in their work, and we came to learn the difficulties they faced in collaborations. 

With these insights, we set out to design a SaaS tool that accelerated time-consuming tasks into smaller multifunctional processes, using a node-based system that evoked the play of building with Lego blocks. We were also inspired by the successes of existing collaborative design platforms, integrating both approaches to the scientific process. It allowed scientists to collaboratively construct innovative hypotheses for novel molecular discoveries using AI and interact with results in flexible ways.

Following the design stage, we began piecing together the proof of concept, built jointly by IBM Research and Experience.Monks. The product necessitated a deep dive into understanding IBM's backend framework, which subsequently guided the design of a bespoke API. We ensured seamless integration with other IBM offerings while catering to frontend needs and proactively addressing potential technical issues. The custom JavaScript SDK for the API now enables the IBM team to deploy it across other IBM Research applications.

By redesigning the research process around creative behaviors, our platform made generative AI accessible to scientists and reimagined accelerated discovery.