Skip to content

Services

Private AI platforms

Open-weight models run in your own EU cloud or on your hardware, behind one gateway you control. You can swap models without changing your apps, and we hand you the weights.
Book an auditDiscuss your model needs

Who it is for

  • Teams hitting cost, latency or privacy walls

    Your current AI stack costs too much, responds too slowly or sends data where it should not go.

  • ML teams moving to owned inference

    You run on third-party APIs today and want to serve your own models.

  • Organisations that cannot send data to third-party clouds

    Your data and keys have to stay on infrastructure you control.

What you get

Running AI reliably in production takes more than prompting.

  • Readiness audit

    A data map, an AI policy, a use-case shortlist and a tier plan.

  • A gateway you own

    One gateway in front of every model, under your keys, so you can swap models without changing your apps.

  • Model selection and benchmarking

    Models compared on cost, latency and quality using your actual tasks. Open-weight first when it wins.

  • Models customised on your data

    Fine-tuned on your documents, tickets, code or transcripts, and adapted to your tone, terminology, compliance boundaries and refusal rules. The weights are delivered to you.

  • Serving sized to your hardware

    Quantisation and serving setup, hardware sizing and a local-first architecture. You own the infrastructure end to end.

  • Inference cost audit

    A token-level cost breakdown of your current stack against API, open-weight and hybrid options.

  • Guardrails and evaluation

    An evaluation harness, guardrails and human approval for high-stakes actions, built into the platform.

  • Post-training operations

    Eval regression checks and quarterly re-tunes as your data grows.

How it works

Every engagement follows the same five steps. Run is the monthly retainer.

  1. 1

    Audit

    Map your data, pick one use case and a metric.

  2. 2

    Pilot

    Live in weeks, measured against real tasks.

  3. 3

    Build

    Production system in your own tenant.

  4. 4

    Handover

    Training, exit pack, every key rotated.

  5. 5

    Run

    Monitoring, model updates, a new use case each quarter.

Sovereignty tiers

From fully on-prem to opt-in frontier models. The audit tells you which tier fits each use case.
Sovereignty tiers
TierWhat it means
T0On-premModels run on hardware in your own premises.
T1Your EU cloudModels run in your own cloud account in the EU.
T2ManagedThe platform is managed for you.
T3ConfidentialModels run on confidential computing infrastructure.
T4Frontier, opt-inFrontier models, used only when you opt in.

Ecosystem and stack

Open models, runtimes, harnesses, labs and hardware we work with.

Packages

  • Readiness audit

    A fixed-scope audit: data map, AI policy, use-case shortlist and a tier plan.

    Book an audit

  • Pilot

    One use case live in weeks, measured against real tasks.

    Discuss a pilot

  • Production build

    A fixed-price production system in your own tenant, handed over with training, an exit pack and every key rotated.

    Discuss your model needs

After handover, the Run retainer keeps the platform monitored, updated and growing.

Questions

Where do the models run?
That depends on the tier you choose, from hardware on your own premises to your own EU cloud account. The audit ends with a tier plan.
Do we get the model weights?
Yes. When we fine-tune a model on your data, the weights are delivered to you.
Can we change models later?
Yes. Your apps talk to the gateway, so you can swap models without changing them.
Do you ever use closed models?
Only when no open model is good enough, and we write down why. Frontier models are the opt-in T4 tier.

Start with a readiness audit.

Map your data, pick one use case and a metric. The audit ends with a tier plan.
Book an auditDiscuss your model needs