Berkeley’s Research IT team has introduced a new AI-powered chatbot designed to assist researchers in navigating the complex Savio supercomputer system, enhancing usability and accessibility across UC Berkeley’s research landscape. This collaborative development effort by Research IT Domain Consultants Noah Baker and Jeffrey Jacob, and Mark Yashar represents a significant step forward in research computing support.
The Research IT chatbot project has been an evolving project over the past year, and addresses a fundamental challenge: while the Savio supercomputing cluster is an invaluable resource, its complexity can present itself as a steep learning curve for users from various disciplines. As Baker notes, "Different departments may use Savio in vastly different ways. A biology researcher's needs might differ significantly from those in physics." While comprehensive, the extensive documentation can be overwhelming for researchers trying to find specific information for their use cases.
The chatbot utilizes a generalized language learning model (LLM) trained specifically on Savio documentation, accessible via a chat interface on the Research IT documentation website. The chatbot's knowledge base encompasses both general computing concepts and Berkeley-specific documentation, allowing it to address a wide range of user needs.
The project has provided valuable insights for the team. Jacob reflects, “The rapid pace of technological advancement has taught us that Research IT must be equally agile in our approach to supporting research.” This experience has underscored the necessity of adaptability to stay ahead of researcher needs.
Baker's focus on ethical frameworks and research needs helped create a solution that met campus requirements while serving the research community effectively. He emphasizes the dual responsibility of embracing new technologies while carefully considering potential risks: "We have a fundamental obligation to provide robust, reliable services to support our researchers." This balance between innovation and responsibility has been central to the project's success.
The chatbot represents more than just a technical solution; it exemplifies Berkeley Research IT's commitment to supporting the campus research community with innovative tools while maintaining high standards of service and reliability. As research computing continues to evolve, this project sets a precedent for how educational institutions can leverage AI to enhance research support services. This tool is designed to provide an enhanced level of support by offering immediate assistance with our comprehensive array of services and documentation, serving as a primary point of contact for technical questions at any time, complimenting our office hours, ticketing, and e-mail help. Available 24/7, the chatbot ensures that researchers have continuous access to guidance and support, helping them get on the right track even outside of office hours. This tool is part of our ongoing commitment to enhancing the accessibility and effectiveness of research support at UC Berkeley.
Users can interact with the chatbot on the Research IT documentation website by clicking the blue circular logo on the bottom right of the screen.