Recently, I was exploring Claude Code alongside with Modelica workflows — and let’s just say… it doesn’t get tired of reading 2,000-line .mo files.

Here’s what changed:

•  Model Understanding – In this case I used Building library. It helped analyze structure, inheritance, parameters, KPIs.

• Automated Script Generation – Python post-processing, batch runs, parameter sweeps — generated quickly in minutes

• Results Processing & Analysis - Generate plots and visualizations from simulation data

• Documentation Generation - Auto-generate documentation from Modelica code

• CI/CD pipelines - model testing and regression validation.

Definitely the good part is it can handle post processing with python efficiently. Of course, it does not validate physics or catch modeling errors (e.g., wrong sign conventions, violated asserts, modeling assumptions). Engineering judgment remains essential.

But this raises a bigger curiosity:

What if we can make use of AI beyond repetitive tasks?

• generation of reduced-order models

• Predictive control (ML-based) for optimization (crucial for giga-scale data center simulations)

• Smarter parameter space exploration