case study — academic project
Bayou Alert
Real-time flood-warning dashboard · Houston, TX · Academic project
One dashboard that pulls scattered flood signals — gauge sensors and citizen reports — into a single operational picture for first responders.
- Role
- Backend & cloud architecture
- Stack
- Azure Functions (Python) · Cosmos DB · React · Mapbox · Tailwind
- Scope
- Academic project
the problem
Houston carries some of the worst flood risk in the US. In a major flood, distress signals scatter across 911 calls, radio, social media, and government sensors — siloed and hard to synthesize exactly when responders can least afford it. A single real-time picture of where water is rising is the difference between fast and delayed dispatch. Bayou Alert is a concept for that picture, built around the region's live gauge network.
what i built
A serverless ingestion backend feeding a React dashboard, so responders can read live gauge conditions at a glance.
Backend — Azure & Python
- Timer-triggered Azure Function polling USGS gauge data from Buffalo Bayou's network on a schedule
- Cosmos DB as the store for alerts, readings, and citizen reports
- REST endpoints serving live and historical alerts plus citizen-submitted reports
Frontend — React + Mapbox
- Responsive dashboard: live map, tables, and charts (Recharts)
- Severity-graded markers across the bayou network
- Dark/light theme for dispatch-center readability
- Sensor and citizen reports aggregated into one view
An NLP pipeline to triage urgent inbound SMS reports is designed but not yet complete — the current build focuses on the gauge-ingestion backend and the dashboard.
the ops angle
This is as much an operations problem as a software one: unattended, scheduled ingestion that can't silently fail; a serverless backend that scales with a crisis instead of buckling during one; and a data layer tuned for fast reads under load. The work spanned the architecture, the serverless backend and data model, and the React client.