We are introducing Strata, a language model we built specifically for government and public-sector work in Indonesia.
General-purpose AI models are plentiful now, and most of them are capable. Capability alone, though, is not always useful. A general model is rarely at home with the regulations, documents, and working habits of Indonesian institutions. Ask it to explain a concept and it does fine. Ask it to trace the legal basis of a policy or draft an official letter correctly, and it often gets it wrong. Strata exists to close that gap.
Trained on Indonesian regulation
Strata is trained on Indonesia’s public body of regulation, from the national level down to regional rules, across many ministries and agencies. From that it learns the hierarchy of laws, the conventions of official documents, and the procedural patterns that general models never pick up. Because regulation keeps changing, we update Strata so it follows those changes rather than freezing at a single training cutoff.
What it can do
Strata reasons across the law, from statutes and government regulations to ministerial and regional rules, with citations that trace back to the source. When two rules appear to conflict, it weighs them using principles such as lex specialis and lex posterior. It reads official documents in a structured way, recognizes signatures, stamps, and seals, and helps draft official correspondence in the proper register.
Run inside the country
Strata can run privately, either on an institution’s own infrastructure or in our facility in Jakarta. Nothing leaves while a request is processed, so the data is not subject to the US CLOUD Act or to foreign legal demands. Like the rest of Epithre’s products, Strata runs entirely on hardware we operate ourselves.
Where it stands
Strata is still in development and not yet generally available. We are glad to talk with institutions that want to follow its progress or help shape where it goes. There is more on the Strata page, or reach us at research@epithre.com.
