Fabrika is a web app for generating synthetic tabular datasets. You define a schema and it produces realistic rows you can preview and export in multiple formats. It combines deterministic generation for structured fields with LLM generation for the messier, context-dependent ones, so the output holds up as test data, demo data, or a stand-in for records you can't share.

Hi! I'm Cris, I built Fabrika! It started because I kept needing tabular data I couldn't actually get hold of: for demos, dashboard testing, trying out new software... Or if I found a synthetic dataset to use it was always something boring like flowers! So Fabrika lets you define the schema yourself and generates from that. Deterministic generation handles the structured fields, and a LLM handles the ones that need some context to look plausible. You can preview the rows before you commit to anything and export in multiple formats. I built it as my MSc dissertation project, and I've spent the time since getting it deployed and usable rather than just demo-able. It's free to try. I'd really like to hear what breaks. Especially if you work with data that has awkward constraints — I suspect there's a lot I haven't accounted for.

The mix of deterministic and LLM generation is clever, I got a messy address column that actually looked like real addresses without me having to write a regex. Wish the schema UI had a way to reuse field types across columns though.

fabrika

Schema-based Synthetic Tabular Data Generation @ Scale

Schema-based Synthetic Tabular Data Generation @ Scale