How I AI How a Founder Uses AI to Improve Hurricane Tracking and Risk Assessment
Mar 24, 2026
Meet Eduardo Siman, founder and CEO of Worldsphere.ai, building computer vision tools to analyze hurricanes and inform risk, insurance, and evacuations. He explains how satellite imagery becomes wind-field data. He discusses the role of scientific rigor, human oversight in high-stakes AI, and how advances in weather AI and rapid prototyping are accelerating impact-driven solutions.
AI Snips
Chapters
Books
Transcript
Episode notes
Turning Satellite Images Into Hurricane Wind Fields
- Worldsphere.ai focuses two tracks: insurance risk intelligence and turning satellite imagery into full hurricane wind fields via computer vision.
- The National Hurricane Center lacked accurate oceanic wind-field estimates, so they trained models to map satellite patterns to wind fields.
Disclose Data, Limits, And Always Include Human Oversight
- Be transparent about what kind of AI you use, what data you train on, and the model limits to avoid harm in high-stakes applications.
- Include human oversight to catch artifacts or impossible outputs and report model error bounds (e.g., historical within ~5 knots).
ERA5 And WeatherBench Fueled AI Weather Progress
- The rapid rise of AI weather began when labs used ERA5's 40-year replay of high-quality reanalysis data to train deep learning models.
- Open benchmarks like WeatherBench accelerated progress by enabling shared scoring and comparisons.
