Speaker
Description
The design of next-generation scientific instruments, such as detectors for the Electron-Ion Collider (EIC), presents a challenging optimization problem involving high-dimensional parameter spaces, competing objectives (e.g., cost and performance), and computationally expensive simulation and reconstruction workflows. Efficiently exploring these complex design spaces motivates approaches that go beyond traditional parameter scans, combining advanced optimization techniques with scalable computing and increasingly intelligent scientific workflows.
AI-assisted Detector Design for EIC (AID(2)E) is an open-source framework developed to address these challenges by integrating modern optimization algorithms, automated scientific workflows, and distributed computing infrastructure. AID(2)E provides a generalizable approach to computationally expensive, black-box single- and multi-objective optimization problems. It enables systematic exploration of design spaces with Bayesian and evolutionary algorithms, and its modular architecture allows complex workflows to be deployed and distributed across local and high-performance/throughput computing environments.
Building on this foundation, AID(2)E promotes the integration of emerging AI technologies, including large language models and agentic workflow capabilities, to assist researchers in developing and managing computational workflows, monitoring optimization campaigns, and interpreting results, while keeping scientists at the center of the decision-making process.
In this talk, we present the AID(2)E framework, its applications to detector design, and ongoing efforts to incorporate emerging AI capabilities into scientific optimization. We highlight how the convergence of machine learning, scalable computing, and AI-assisted workflows can help design the next generation of scientific experiments.