2026-11-04 –, Bataglieri
Build a serverless vector tile API with DuckDB in Azure Functions, querying GeoParquet directly from Azure Blob Storage. This session outlines the architecture, dynamic tile generation workflow, and a lightweight, scalable alternative to databases and tile subscription services.
The session outlines how to build a tile delivery API using DuckDB, GeoParquet, and Azure Functions+Storage. It will focus on large data sets (100+ million rows) that cross thousands of GeoParquet files requiring a number of indexing, filtering, and pruning approaches to ensure sub 100ms response times for tile generation. All aspects will be covered including GeoParquet file creation optimization (file sizes, row group sizes, geographic grouping), query optimization within DuckDB, and API design. While this will focus on utilizing Azure cloud resources, the techniques are generic and would work in approaches utilizing entirely FOSS like FastAPI and local storage.
My name is Jesse Amundsen. My main focus has been solving complex geospatial problems for the last 15+ years mainly within the telecom industry. I initially worked on spatial programming for network cost models developed for clients in both the private (ex: AT&T, Verizon) and public sector (ex: FCC, State Broadband Offices). I also worked on projects related to auction models and support programs (ex: CAF II, MF II).
After a decade using locational datasets not targeted narrowly enough for the purpose of broadband reporting, I led the technical development of the Broadband Serviceable Location Fabric (BSLF) (Reference #1, Reference #2). The BSLF has been implemented by agencies such as the FCC and NTIA for the purposes of broadband reporting and is used in multiple federal funding programs (ex: BEAD).
Along my journey I have become highly competent in PostGIS, using it to implement the majority of the geoprocessing pipeline that generates the BSLF. Additional areas of strength include software architecture and HTTP APIs, each with a focus on understanding and implementing in a spatial context.
I enjoy thinking through geospatial problems, spatial data science challenges, and advocating for building long-lasting complex systems out of simpler constituent parts.