Nikita Chetverikovzedbyl.tech
Book intro call
Engagements2026

Four engagements. One principle: AI stays inside your perimeter.

A defined progression from architectural assessment to a production sovereign-AI environment operating entirely within client infrastructure. Each engagement is scoped, fixed-fee, and structured so that IT, finance, and compliance can review the commitment before it begins.

01 · Engagement models

Select the entry point for your environment.

Most organisations begin with an architectural assessment, then proceed to deployment. The assessment fee is credited against the deployment engagement.

Note - compute is procured directly by the client against the assessment specification. No subscriptions, no per-seat licensing, no usage metering.

02 · Process

From first engagement to production environment - typically inside three weeks.

  1. STEP 0130 min · confidential

    Qualification

    A confidential review of the organisation, the workflow, and the regulatory posture to confirm whether sovereign AI is the appropriate response.

  2. STEP 0290 min · scoped

    Assessment

    Workflow review, infrastructure sizing, architectural recommendation, written roadmap. Credited against the deployment engagement.

  3. +30 days free support
    STEP 037 working days

    Deployment

    Provisioning, isolation, document ingestion, knowledge transfer. The environment enters operational use within the same week.

  4. STEP 04Monthly

    Operations

    Continuous model lifecycle management, monitoring, and governance reporting under a defined SLA.

03 · Architecture

An open-architecture stack, owned and operated inside your perimeter.

On-premise compute

Inference layer

  • - Open-weight foundation models
  • - Quantised for on-prem hardware
  • - Selection guided by workload and isolation requirements
Document grounding

Retrieval layer

  • - Vector index over the internal corpus
  • - Permission-aware retrieval
  • - Citation back to source documents
Internal automation

Workflow layer

  • - Event-driven internal pipelines
  • - Integration with line-of-business systems
  • - Auditable execution trail
Compliance posture

Isolation perimeter

  • - Outbound egress constrained at the firewall
  • - Air-gap-capable architecture
  • - Audit-ready network policy
Hardware

Compute substrate

  • - Low-power inference nodes for constrained-footprint deployments
  • - GPU substrates where throughput envelope requires
  • - Sized against assessment specification
04 · Productised entry

A pre-commissioned sovereign-AI appliance - Sanctum Box.

A pre-configured on-premise environment delivered with private inference and retrieval, sized to the engaging team. Specification, commissioning, and delivery handled under a single fixed fee.

  • Sized to the engaging team

    Three capacity bands aligned to team size: 2–5, 5–15, or 15+ users. Specification matched to workload at the point of order.

  • On-premise by architecture

    The appliance resides within client premises. No data leaves the perimeter at training, retrieval, or inference time.

  • Operations included

    Model lifecycle management, monitoring, and priority support included under a single engagement.

Best fit if

Sovereign AI is required quickly, without a separate architectural assessment. Team of 2–15+ users, with a defined budget envelope.

Sanctum BoxRequest discovery

Requirements outside the standard configurations are addressed through the Architectural Assessment engagement above.

05 · FAQ

Questions reviewed before engagement.

Why move AI on-premise in 2026?

Regulation, model capability, and hardware economics have converged. The EU AI Act and ISO 42001 have raised the audit threshold for cloud AI in regulated environments. Open-weight models have closed the capability gap with proprietary cloud offerings. A single on-premise node now serves a 70B-parameter model to a working team without per-seat subscription costs. Retaining data inside the perimeter is now a tractable architectural decision rather than a cost trade-off.

Is the environment genuinely private?

Yes. Inference and retrieval execute on client hardware, inside the client perimeter. No outbound call is made to a vendor API at inference time. Network egress is constrained at the firewall by default, and the environment operates under enforced network isolation.

Why not adopt an enterprise cloud-AI product?

Enterprise cloud AI is a capable product. The structural constraint is unchanged: client documents continue to traverse a third-party environment. For attorney-client privilege, patient data, or trade secrets, this is an architectural question, not a contractual one. An on-premise deployment removes the question by removing the external system from the data path.

What is the typical engagement duration?

5 to 14 working days from agreement to handover. The majority of the engagement is spent on document corpus integration and connection to existing internal systems, not on the inference layer itself. Standard commercial terms are 50% on engagement, 50% on handover.

Is a dedicated internal owner required to operate the environment?

No. The architecture is deliberately conservative and open-source throughout. Operational ownership transfers to client IT under a written runbook. Engagement experience indicates most clients require no further intervention for six months or more following handover.

What if the consultancy becomes unavailable?

Every component delivered is open-source and documented. A written runbook is provided at handover. A competent internal administrator can assume operations directly, and a small network of independent consultants operating the same architecture is available as an alternative. No proprietary lock-in is introduced at any layer of the architecture.

Scope your engagement.

A 30-minute confidential discovery to review your environment, regulatory posture, and intended use cases. The session confirms fit and identifies the appropriate entry point. No fee. NDA on request.

Request discoveryBack to home