2026-11-03 –, Bataglieri
DuckDB has become the open-source engine for fast local analytics, with AI agents increasingly building analytics workflows around it. This talk covers using DuckDB, H3, Overture, and open-source tooling to build scalable spatial analytics pipelines and interactive dashboards.
DuckDB is rapidly becoming the FOSS tool of choice for fast, local analytics. AI agents are increasingly being used to develop analytics and dashboards using DuckDB. In this talk, we will discuss our experience on using DuckDB, extensions like h3-duckdb, and our open source data workflow tools to perform interactive data analytics.
H3, an open source project originally developed by Uber, enables efficient entity conflation through fast spatial indexing for identity matching and deduplication, especially when combined with DuckDB for scalable local joins. This pairs naturally with Overture data and Overture’s Global Entity Reference System (GERS) and reduces reliance on brittle string matching and custom reconciliation pipelines.
We will describe how we preprocess and partition Overture data, develop SQL analytics pipelines, and create interactive and beautiful dashboards. We will highlight how open source data workflow tooling orchestrates AI agents to generate code while ensuring transparency.
Isaac Brodsky is co-founder and CTO of Fused, building AI workflows for data. He was previously co-founder of Unfolded, and previously software engineer at Uber on Marketplace Data. He maintains the H3 open source library and the DuckDB extensions h3-duckdb and duckdb-zipfs.
