They all write, store and read data: how to pick the right database ·
SQLite, Postgres, DuckDB and 20 others: they all write, store and read data. The questions you ask every day tell you which one you need.
Data-Engineering
-
-
Microbatch: how to supercharge dbt-duckdb with the right incremental model ·
Learn when and how to use dbt's microbatch incremental strategy with DuckDB. Covers row groups vs partitions, benchmarks comparing full refresh, merge, delete+insert, and microbatch strategies, plus configuration tips and common pitfalls.
-
LLMs Are Cheap Enough to Change How You Work With Data ·
I processed 60,000 job postings with an LLM for $10. That's not a typo. This post walks through how to use OpenAI's Batch API for bulk document processing: building JSONL request files, orchestrating without losing your mind, controlling costs with token limits, and tracking batch state in a database. When AI becomes this cheap, it changes what's possible.
-
DuckDB is my new Excel ·
I'll admit it: as a data engineer I still use Excel. But DuckDB has become my secret weapon for ad-hoc data exploration. No more struggling with CSV imports, no more manual column matching across files, no more giving up on JSON. This guide shows you how to query remote CSVs, handle schema drift across multiple files, unnest nested data, and flatten JSON APIs—all with SQL you can remember if you can remember Excel formulas.
-
Turn Thousands of Messy JSON Files into One Parquet: DuckDB for Fast Data Warehouse Ingestion ·
If you've inherited a bucket full of thousands of tiny JSON files—one per API call, one per event, one per log minute—you know the pain: slow scans, schema anxiety, and rising warehouse bills. This guide shows you how to consolidate them into clean Parquet with DuckDB: handling schema drift, maintaining lineage, optimizing performance, and integrating with dbt. Touch your raw files once, then model against something stable.
-
Beyond Tables and Views: Building a Custom dbt Materialization to Deploy Streamlit Apps on Snowflake ·
What if dbt could deploy a dashboard instead of just another table? I was tired of switching between my dbt transformations and Streamlit visualization code, so I built a custom materialization that deploys Streamlit apps directly on Snowflake. This tutorial walks through the architecture: using dbt configs, uploading files to stages, and creating apps that know about your data models, lineage, and freshness.
-
Unit Testing dbt Macros: A workaround for dbt's unit testing limitations ·
Remember when dbt projects meant folders full of SQL files and crossing your fingers that your transformations were correct? dbt 1.8 finally brought unit testing, but macro testing is still limited. This post walks through unit testing your business logic—like standardizing campaign sources—and a workaround for testing macros so you can catch broken SQL before it wrecks your dashboards.