Simulation workflow agent for carbon storage

Lately I have been thinking about what it would look like to build a simulation workflow agent for carbon storage.

This idea comes from a very practical place. In reservoir engineering and simulation work, a lot of time goes into the steps around the actual model run: checking assumptions, preparing inputs, fixing errors, reading outputs, comparing cases, and deciding what to try next. Each step may look small by itself, but together they shape the pace of research and engineering work.

So the question I keep coming back to is: can an AI agent make this loop smoother?

I do not mean an agent that magically replaces expertise or makes decisions on its own. I am more interested in something that works alongside the researcher or engineer, helping connect documentation, code, simulation tools, results, and human judgment. If the agent can reduce friction in the workflow, even a little, that could be valuable.

Two things I read recently made this idea feel more concrete:

• JutulGPT - Agent-based reservoir simulation with JutulDarcy.jl by Olav Møyner
• 24/7 Simulation Loops: How Agentic AI Keeps Subsurface Engineering Moving by NVIDIA

I am still at the early thinking stage, but I would like to start exploring this more seriously. If you are interested in carbon storage, reservoir simulation, scientific machine learning, or AI agents for engineering workflows, I would be happy to connect and look for possible collaborations.