eatmydata.ai - answer questions about your data without exposing it to AI

We automate daily business problem - you have your data sitting next to you, and you want to answer a random business question to support your thinking, ideation, and decision-making. Save the answer to re-evaluate later on fresh data, or share it with a colleague.

"How much money did lost due to sneakers warranty claims related to torn or faulty shoelaces?" - good luck constructing an SQL of Excel formula to find this out.

We answer questions likes that under 10 seconds, end to end, presenting you answer, charts (if needed), all data used and a non-technical explanation what is what (if you drill down to it).

When we do it, we don't disclose your data to LLM, while at the same time still using excellent coding agents in realtime to build, run and show you a one-off application to analyze and answer your question, all in under 10 seconds!

Under the hood, we:

• Combine SQLITE OPFS sync handle and WASM SQLite for blazing-fast local data querying;

• Use realtime NER and Bayesian models to prevent data leakage to remote LLM;

• Employ latest advanced in semantic vector search such as TurboQuant (bundled in our sqlite) and Model2Vec for realtime embeddings building;

• Sanboxed QuickJS for AI code execution;

• An `orcherstator -> sql_planner -> js_coder` agent loop to write sql+js produce structured dashboards;

• Apache Echarts for beautiful dashboard rendering;

And it's open-source and MIT-licensed: https://github.com/eatmydata-org/eatmydata