How Noolio handles your data
Noolio runs Stata on your machine. Command writing and error correction run in the cloud. This page covers what moves between the two.
- Your datasetStays local
- Stata
- Local Python backend
.do file- Stateless proxy
- LLM providers
What stays on your machine
A local Python backend talks to your licensed Stata install.
- Your datasets (
.dta,.csv, Excel) stay on your hard drive. - Data cleaning, modeling, and statistical computation run on your hardware.
- Noolio's servers do not receive your dataset file.
What goes to the cloud
To turn your instructions into Stata code and fix errors, Noolio sends analytical context to our backend. The backend is a stateless proxy. We do not store session data. The diagram above itemizes exactly what reaches our LLM providers.
Which models we use
Noolio routes requests across several models, pairing faster planning models with heavier code-writing ones. The routing engine currently uses Anthropic Claude, OpenAI, Google Gemini, and GLM.
If you work with PHI or regulated data
Because Noolio relies on Stata log outputs to verify execution and debug syntax, it is possible for snippets of your data to be included in the transmitted logs (for example, if a list, tabulate, or summarize command echoes real values into the Stata console).
If you work with Protected Health Information (PHI) or strictly regulated institutional data, please be aware that these log snippets are transmitted to our cloud LLM providers to facilitate the coding loop.
For local-LLM deployments and on-premise compute cluster (HPC) integrations for clinical and strictly regulated research environments, contact us to discuss institutional deployment.