FloodIQ

Inspiration FEMA's flood maps are outdated in 75% of cases — yet every homebuyer relies on them to make an $800,000 decision. After Hurricane Harvey, half the homes that flooded were outside FEMA's designated hazard zone. We wanted to build the tool that gives regular people the same risk analysis that insurers and developers already have but never share.

What It Does FloodIQ takes any US address and returns a hyperlocal flood risk score built from real physical data — Scripps Institution sea surface height trends, NOAA elevation, and FEMA historical disaster records. It places that score directly alongside the official FEMA designation so the buyer can see exactly where the government's map and the physical reality diverge, with a plain-English explanation of why.

How We Built It We built FloodIQ on a four-layer architecture. AWS Lambda and API Gateway handle live data ingestion — pulling FEMA zone designations, USGS elevation, and historical flood records on demand. Databricks processes the Scripps 40-year sea surface height simulation into a queryable Delta Lake risk surface. A scoring model trained on FEMA disaster declarations is deployed via Amazon SageMaker as a real-time inference endpoint. The frontend is built in React and deployed on Vercel.

Challenges We Ran Into The Scripps SSH dataset is a 40-year NetCDF simulation — a format none of us had worked with before. Learning xarray under time pressure while simultaneously wiring four independent layers together was the core technical challenge. Getting CORS headers right across Lambda and API Gateway so the frontend could actually call our endpoints without errors cost us more time than we expected. Connecting Databricks to S3 also required IAM permissions we had not anticipated needing.

Accomplishments That We're Proud Of We built a genuinely working end-to-end pipeline in a single hackathon — four independent layers that each person owned completely, merged cleanly in under two hours. More than the technical execution, we are proud that the product addresses a real and documented injustice: millions of Americans are making the largest financial decision of their lives with information that every other party in the transaction knows is wrong.

What We Learned That the hardest part of a multi-person technical project is not the code — it is defining the contracts between people before anyone writes a line. We learned that deploying dummy endpoints early and building against fake data in parallel is the only way to avoid blocking each other. We also learned that NetCDF files and oceanographic data are far more accessible than they look once you understand the xarray data model.

What's Next for FloodIQ The immediate next step is replacing the weighted scoring function with a fully trained model incorporating more granular climate projection data. Beyond that, the product has a natural expansion path — integrating wildfire risk, heat island exposure, and sea level rise projections into a single unified climate risk score for any property. The long-term vision is becoming the climate risk layer that sits inside every real estate transaction in America, giving every buyer what only insiders had before.