DataSpoc is an open-source data platform built for simplicity.

  Most data teams drown in infrastructure before they query a single table. DataSpoc takes a different approach: three CLI tools that communicate through Parquet files in a cloud bucket. No
  Airflow, no Spark, no warehouse.

  Pipe — Data Ingestion (Open Source)
  Connects any of the 400+ Singer data sources to Parquet files in S3, GCS, or Azure. Streaming conversion, incremental extraction, auto-catalog. One command to run a pipeline.

  Lens — Virtual Warehouse (Open Source)
  Mounts your cloud bucket as SQL tables powered by DuckDB. Interactive shell with autocomplete, Jupyter and Marimo notebooks, AI natural language queries, and local cache for offline work.
  No server, no infrastructure.

  ML — AutoML (Commercial)
  Train, predict, and explain machine learning models directly on your data lake. Feature engineering, model selection, and deployment — all from the command line.

  The architecture is deliberately simple. Pipe writes Parquet. Lens reads it. ML consumes and produces. The bucket directory convention is the only contract between them. All access control
  is delegated to your cloud provider's IAM.

  Built in Brazil. Python. Apache 2.0 (Pipe & Lens).

  GitHub: github.com/dataspoclab
  PyPI: dataspoc-pipe, dataspoc-lens

DataSpoc

Org chart

Teams

Offices