The question
Much of the data that could provide early warning for disasters already exists. Population movement shows up in call detail records, environmental stress in satellite imagery, and vulnerability in ground surveys. But it lives in different formats, resolutions, and institutions. Can we fuse it into systems that give humanitarian responders lead time?
The system
At CIDER (Cornell Institute for Digital Engagement and Resilience), I’m co-authoring a white paper for Cornell’s “Data Science to Build Resilience and Improve Humanitarian Response” Thought Summit. The technical core is multi-modal data fusion for disaster early warning.
What I built
- CDR features that turn raw call detail records into mobility and displacement signals robust enough to serve as model inputs.
- Satellite imagery pipelines for automated ingestion and feature extraction of environmental stress indicators.
- Fusion architecture that aligns sources which differ in time, space, and geographic units.
Beyond the models
Half the work is translation. I write the policy narratives that carry model findings to decision-makers and engage the summit’s funders and partners directly, because the models only matter here if institutions act on them.