Siftbase
Turns messy file uploads into schema-aligned datasets, with AI-drafted pipelines you review before they run
Project Overview
Siftbase takes source files — CSV, XML, XLSX — and turns them into reviewed, schema-aligned datasets. You upload the files, pick or create a target schema, and the AI proposes a transformation pipeline. You review the SQL, refine it, run it, and apply the output to a published dataset.
The point is the review step. Plenty of tools will transform your data for you; the hard part is trusting the result enough to publish it. So nothing is a black box — you read the SQL before it runs, see how confident the model is, preview the output, and roll a run back if it wasn’t what you wanted.
Stack
Python, DuckDB, React.
Currently MVP, working with early design partners.