Julius

Julius is a data analysis assistant founded in San Francisco in 2022 by Rahul Sonwalkar and operated by Caesar Labs. Upload a spreadsheet, CSV or other dataset, ask a question in plain language, and it writes and runs the analysis code, returning the result as a chart, table or statistical output with the code visible. It targets the researcher who knows what analysis they want but does not want to write the Python or R to get there.
Why it matters
The gap between having a dataset and having a first look at it is mostly mechanical, and it is where a lot of research time goes.
Showing the generated code makes the analysis auditable, a black-box chart would be unusable for anything that has to be defended.
It lowers the barrier for wet-lab and qualitative researchers who need statistics but do not program.
Fast iteration over a dataset changes what gets explored, questions that were not worth the scripting effort get asked.
Key features & advantages
Natural-language analysis over uploaded datasets, with the generated Python visible and editable.
Charts and visualisations produced directly from the query, exportable for reports and figures.
Statistical modelling and forecasting, including regression and standard hypothesis tests.
Report and notebook export for handing an analysis on.
Limitations & things to watch
Generated code runs without error and still answers the wrong question, the code is shown for a reason and needs reading.
Statistical test selection is inferred from the phrasing of the request, an inappropriate test will be applied silently if that is what the question implies.
Uploading data to a hosted service is a governance decision, check it against your data agreements before any sensitive or patient-derived dataset goes near it.
Results are not automatically reproducible as a pipeline, export the code if the analysis has to be re-run or published.