Valleru, Lohit; Newlin, Nancy R.; Girija Paramasivam, Neeraj Harikrishnan; Brophy, Erica; Cheong, Jamie S.; Krauskopf, Alejandro A.; Levine, Jonathan A.; Pulido, Iván; Sahin, Merve; Chakradeo, Monica. (2026).Ìý.ÌýPEARC 2026—Practice and Experience in Advanced Research Computing 2026: Resilient Roots + Empowered Communities.Ìý
Cancer researchers increasingly rely on high-performance computing (HPC) to run powerful but complex computational tools. However, researchers without specialized computing expertise can face difficulties setting up software, managing data, submitting jobs, and monitoring results. To address these barriers, Memorial Sloan Kettering Cancer Center developed Iris AI, a chatbot that allows researchers to interact with its HPC systems using everyday language. Using AlphaFold3, a tool for predicting protein structures, as an example, Iris AI enables researchers to prepare inputs, launch and monitor GPU-intensive analyses, and view results through conversational commands. A preliminary evaluation with five researchers compared Iris AI with the standard command-line workflow. The system simplified several technical steps, making advanced computational tools more accessible and potentially improving the efficiency and reproducibility of cancer research.
Figure 1.
Iris AI provides a unified tool that integrates the Claude Sonnet (version 4.6), Open OnDemand user interface, and agents that interface with MSKCC’s HPC cluster and an AlphaFold3 container. It supports queries for navigating the on-premises cluster, creating and formatting files, submitting AlphaFold3 jobs, and rendering CIF files in an embedded interactive viewer.