Regnant is a Sovereign Artificial Intelligence research and infrastructure company, founded in Tanzania. The prevailing AI stack was never designed for the sovereignty, languages, or decisions of emerging institutions. We are building the alternative, deliberately, and to a standard institutions can stand behind.
Research & infrastructure
Owned, not rented
We concentrate on three fronts, and decline to spread thinner than we can stand behind.
Sovereign language models
Models built for the languages institutions actually operate in, engineered for precision and accountability, not approximation.
Decision simulation
Deterministic systems that let institutions rehearse consequential decisions, and understand their downstream effects, before committing to them.
Sovereign infrastructure
Private inference, semantic grounding, and the runtime that keeps every layer of the stack within the institutional perimeter.
Founder · FDR.00
Dar es Salaam
The institutions that carry the most responsibility were handed the least control.
They were told to rent their intelligence, route their data through infrastructure they do not control, and trust that the terms would remain fair. That is not a foundation a country can build on.
So we build the opposite: systems an institution owns down to the weights and the hardware, models that operate in the languages people actually govern in, and tools that can account for every decision they influence. We would rather ship fewer things that are genuinely sovereign than many that quietly depend on someone else.
This is early, deliberate work. If it resonates, I would welcome the conversation.
Daudi.
the cloud cannot fix
When the cloud is the adversary.
Every request routed through foreign infrastructure reveals something about what an institution is monitoring, planning, or concerned by. Running inference inside your own perimeter removes that exposure entirely.
When the model cannot read the language.
A decision-maker briefed by a system that cannot reliably work in their working language is constrained by the tool, not served by it. In governance, language proficiency is not optional.
When the decision cannot be undone.
Infrastructure, procurement, public-health response: decisions whose consequences unfold over decades. The responsible posture is to model the alternatives before committing to any one of them.
01Sovereignty before convenience.
When ownership is harder to deliver than access, we choose ownership. Control over critical infrastructure is the objective, not a concession to be traded away for ease of use.
02Language is infrastructure, not a feature.
A system that cannot operate in the language an institution governs in does not serve that institution. We treat linguistic fluency as a first-order requirement, engineered from the outset.
03Systems that govern must explain themselves.
Every consequential decision our software informs must be reconstructable, auditable, and defensible. Where decisions carry public weight, there is no place for opacity.
04We ship only what we can account for.
We prefer a smaller portfolio we own and understand completely to a broad one that quietly depends on infrastructure we do not control.
Sets the technical vision and institutional strategy: the thread connecting language equity, decision autonomy, and institutional sovereignty.
Leads sovereign compute: the air-gapped inference clusters and isolation infrastructure that run entirely within an institution's own perimeter.
Leads the language mission: the pipelines and evaluation frameworks that let our models work in under-served languages with real fluency.
Owns field deployment: turning architecture into running systems inside finance ministries, central banks, and state operations centres.
Built here. For everywhere.
We are headquartered in Dar es Salaam because the problems we work on are most visible from here. The solutions apply wherever institutions need intelligence infrastructure that answers to them, and no one else.