fantine/microseismic-detection-ml — reverse-engineered prompt

Build me a reproducible machine learning project for detecting microseismic events in fiber optic DAS data.

I want it to take raw or prepared seismic data, convert it into TensorFlow records, train a deep learning model, run inference on continuous data, and save detections and logs so I can review what events were found. It should be set up for geophysics research use, with config files for data paths and model settings, scripts to run training and prediction jobs, and clear docs explaining the workflow.

Please include support for running everything in a GPU container, with notes for Docker and Singularity. I also want a simple hyperparameter tuning workflow and a modular structure so I can swap settings without rewriting code. Keep it practical and command line based, since this is more like a research pipeline than a web app. Look up current docs online if you need to.