AdI145-12/MatRisk-AI-V2 ? reverse-engineered prompt

Build me a Python app called MatRisk AI that ties material science and financial risk together in one working project.

I want it to read the datasets from a local extracted folder, train the core models, and produce the outputs described in the README, like material property predictions, physics checks, commodity signal backtests, infrastructure credit risk scores, insurance catastrophe risk, ESG summaries, inverse design candidates, and the MatRisk Lab game results. Please make the main script easy to run, keep the code modular, and add tests so the main pieces work without needing the real data files.

Also set it up so I can run the dashboard in Streamlit, and use Docker and the provided configs if they help. If anything in the current docs is unclear, look up current package docs online and make sensible choices. I’d like it to feel like a polished demo project that runs end to end and saves results into the output folder.