// Software & Data · July 2025
Storm Formation Analysis Tool
Pulls satellite imagery and live weather data together to look for the conditions that precede a storm.
The Problem
Storm formation shows up in two kinds of data that don't naturally line up: what the cloud field looks like from above, and what temperature, pressure and humidity are doing underneath it. Neither on its own says much, and they arrive in different formats from different providers.
The Approach
Built a Python pipeline that ingests both and treats them as one dataset — satellite imagery for cloud-formation features, the OpenWeatherMap API for the numerical environmental picture — with feature extraction over the imagery aimed at spotting cumulonimbus development early. Kept the stages modular so a model could be dropped in later, rather than building prediction in from the start.
The Outcome
An ingestion and analysis pipeline rather than a finished predictor. It reads both sources and extracts features, but there is no trained model behind it and no alerting, so it surfaces patterns rather than forecasting from them. The honest limitation is that the machine-learning step the pipeline was shaped around was never added.