Tabkeel is built by Francisco Ferreira, a founder and engineer who kept getting burned by the same thing: discovering that a number was wrong after a decision had already been made on it. The pipeline had quietly stalled, a table had filled with nulls, a metric had doubled — and the dashboard said nothing, because dashboards wait for you to look.
Francisco spent seven years in data and machine learning, including at companies in the financial sector — building RAG pipelines, LLM agents and production evaluation systems. He also builds PromptEval, a prompt-evaluation platform. Tabkeel is the tool he wanted on the data side: rigorous under the hood, simple on the surface.
Most data-quality tools assume you already have a data team. We didn't, and neither do the founders, PMs and full-stack engineers we build for. So Tabkeel takes the opposite stance: connect read-only in two minutes, let it learn what normal looks like for each table, and get told in plain language when something drifts — with the query to diagnose it already written.
We run our own company on it. Tabkeel watches the production database of PromptEval, our other product, every day. The alerts you see on the marketing site are the kind we actually get on our own data — not mockups.
Two principles don't move: your database stays read-only, and you stay in control of what Tabkeel can see. We'd rather earn trust slowly than ask for it.
Questions or feedback? francisco@tabkeel.com.