Leveraging LLMs for Automated Post-Processing of HPC Simulation Output Logs

James Imrie- Senior Solutions Architect

Engineering teams are generating terabytes of CAE and HPC simulation data every day, but extracting reliable, actionable insights from solver logs is still a manual, error‑prone bottleneck. This session dives into how Large Language Models (LLMs) can be embedded directly into simulation workflows as a “co‑pilot” for post‑processing and analysis.

You’ll see how local, enterprise‑grade LLMs can:

• Parse and interpret complex solver logs across tools like Siemens Star‑CCM+, Abaqus, and Ansys Fluent
• Automatically classify errors and failure modes (mesh quality, divergence, memory, licensing, convergence issues)
• Extract and normalize key simulation results to build AI‑ready datasets from trusted runs
• Power automated troubleshooting flows that reduce support load and increase user autonomy
• Operate in isolated, private environments with CI/CD, benchmarking, and regression testing for production use

The session walks through the full lifecycle of deploying LLMs for simulation post‑processing: model selection and benchmarking, data pipelines, prompt design, CI/CD integration, and validation against customer‑specific use cases.

Watch this session to see real-world LLM deployments on production solver logs, understand the reference architectures behind them, and leave with patterns you can apply immediately to automate error handling, speed up root-cause analysis, and turn simulation output into an AI-ready data asset.