MapWork.ai

Inspiration

Commercial roofing companies spend countless hours searching for potential customers by manually browsing maps, driving through industrial areas, or purchasing expensive lead lists that quickly become outdated. I wanted to explore whether publicly available geospatial data and satellite imagery could automate this process and help contractors focus on the buildings that matter most.

What it does

MapWork.ai is an AI-powered geospatial platform that identifies commercial flat-roof buildings and transforms them into qualified sales opportunities.

The platform combines OpenStreetMap building footprints with Mapbox geocoding services to locate commercial properties, enrich them with address information, and store everything in a PostgreSQL/PostGIS database for efficient spatial querying. Satellite imagery is then linked to each property, forming the foundation for future AI models that can detect roof damage and prioritize buildings that are most likely to require maintenance.

How we built it

The backend was built using Python and Flask, with PostgreSQL + PostGIS handling geospatial queries and indexing. OpenStreetMap and Mapbox APIs provide building footprints and reverse geocoding, while AWS S3 stores satellite imagery for future analysis.

A major focus was performance. By optimizing spatial indexes and database queries, the platform can retrieve geospatial data in under 300 ms, making it scalable for thousands of commercial properties.

Challenges we ran into

One of the biggest challenges was working with large geospatial datasets. Building footprints often contained inconsistent or incomplete information, requiring multiple filtering and validation steps before they became useful leads. Efficiently querying spatial data while keeping response times low also required careful database design and indexing.

Another challenge was designing the system with future AI integration in mind. Rather than building a one-off application, I wanted an architecture that could easily incorporate computer vision models for automated roof damage assessment.

What I learned

This project taught me that building an end-to-end product is much more than writing code. I learned how backend systems, geospatial databases, cloud storage, and external APIs work together to solve real-world problems.

What I'm proudest of isn't the technology itself—it's seeing local roofing companies express genuine interest in using something that started as a simple idea. Those conversations reinforced that the most meaningful software isn't defined by its tech stack, but by the value it creates for real users. That experience continues to motivate me to keep improving MapWork.ai and exploring how AI can make traditionally manual industries more efficient.

Built With

• amazon-web-services
• api
• backend
• docker
• flask
• geocoding
• imagery
• mapbox
• openstreetmap
• postgis
• postgresql
• python
• remote
• rest
• reverse
• s3
• satellite
• sensing
• sqlalchemy