Cohorts forming for the 2026 intake — dates and fees confirmed before you commit.

Register interest →
Skip to content

Sovereign AI: Local Models, Data Residency & National AI Stacks

Four weeks on running capable AI inside a boundary you control: selecting and serving open-weight models, quantisation, routing between local and frontier models, and the residency requirements that drive the architecture.

  • 4 weeks
  • Live online
  • Advanced
  • English
  • Dates announced soon
Join the waitlist

Who should take this program?

You'll get the most from this if

  • Confident Python
  • Comfortable with containers
  • Some LLM experience

This is not for

  • Anyone expecting a policy course — this is engineering

What you will build

  • A locally served open-weight model with measured throughput and cost
  • A routing layer choosing between local and frontier models by task
  • An honest capability comparison on your own evaluation set

What you learn, week by week

Module 1 — What sovereignty actually requires

  • Separate legal requirement from preference
  • Map which workloads can leave the boundary
  • Cost the alternatives honestly

Module 2 — Serving open-weight models

  • Run and benchmark a model locally
  • Apply quantisation and measure the quality cost
  • Size hardware against real throughput

Module 3 — Routing

  • Build a router that chooses by task and sensitivity
  • Fall back safely when the local model is not enough
  • Measure the capability gap on your own evals

Module 4 — Operating it

  • Handle upgrades and drift without a vendor
  • Monitor a self-hosted stack
  • Present the total cost of ownership

How this program handles evaluation and governance

Sovereignty claims are checkable or they are marketing. The course insists on the checkable version: can this workload run at acceptable quality inside your boundary, and can you prove where the data went? You will measure the capability gap rather than assume it, in both directions.

Tools and stack you will use

  • Ollama
  • vLLM
  • Open-weight model families
  • Docker
  • Python
  • OpenTelemetry

Fees

Request pricing

Fees depend on cohort, format, and whether this runs for an individual or a team. We'll send the full breakdown.

Get fee details

Questions people ask

Do I need a GPU?

For part of it. We provide lab access if you do not have hardware.

Is self-hosting cheaper?

Sometimes. Break-even depends on volume, and teams get this wrong in both directions. You will calculate it properly.

Does this cover UAE and India residency law?

We cover what the requirements mean architecturally. Legal interpretation belongs with your counsel.

Ready to build AI that holds up under scrutiny?

Join the waitlist
Join the waitlist