Agentic AI for Science and Engineering

Developing autonomous AI agents to automate complex scientific workflows—from simulation to analysis—allowing researchers to focus on decision-making rather than process management.

Status: Active

Agentic AI for Science and Engineering focuses on the work that sits between a scientific question and a reliable result. In most projects, researchers spend weeks rebuilding the same workflow: connecting scientific software, preparing data, running simulations or experiments, fitting models, and debugging failures before they can test an idea.

This work aims to address that bottleneck by developing AI agents that take an objective, plan the steps, and execute the workflow across simulation, analysis, and experimentation through well-defined interfaces to tools and compute resources. The agent monitors progress, runs checks grounded in physical constraints and uncertainty, and asks for human input when choices trade off cost, risk, or evidence quality. The goal is to reduce time spent keeping workflows running and increase time spent interpreting results and making scientific decisions.