We built a serverless data stack from the ground up using DuckDB, SQLMesh, and S3 for a global nonprofit. Our open-source-first approach eliminated SaaS dependencies while empowering developers with reproducible environments, Python-powered transformations, and collaborative workflows.