vanwebdev product visual: Consulting Services

Sovereign Infrastructure Brief

Starting at $3,000 CAD

Hardware, network, and model architecture for production AI. Range: $3K to $10K.

Book a Kickoff Call Download Engagement Brief

Scope of engagement

What you get

  • One AI workload per brief. We focus rather than survey, and the topology serves that workload.
  • GPU and CPU AI workloads scoped end to end. We size the compute against your throughput envelope.
  • Open-weight model hosting on Canadian compute. We name the model, the host, the jurisdiction.
  • API model proxying with egress controls. We map each external call to the byte-level boundary.
  • Retrieval index design for Canadian data corpora. We pick the index against your latency budget.
  • Vector store vendor choice with jurisdiction filters. We rank vendors by Canadian-region availability.
  • Observability stack with audit-trail retention. We name the retention window your auditor demands.
  • Key management with HSM or KMS shortlist. We name the key custodian and the rotation cadence.
  • Egress map and data-residency boundaries. We trace every byte; we mark every country code.
  • ITSG-33 control family mapping per workload. We cite the family; we name the control number.

Timeline

3 to 5 weeks

Deliverables

  • Sovereign topology diagram: A one-page topology diagram of the proposed sovereign AI stack for the workload you bring to the call. On-prem racks, private cloud, or hybrid topology. We name every node. We draw every interconnect. The diagram ships as PDF the same week and as an editable source file for your architects.
  • Canadian-first vendor shortlist: A vendor shortlist with Canadian-jurisdiction options named first across six layers: GPU compute, model hosting (open-weight and API), retrieval indices, vector store, observability, and key management. Specifically, each named vendor carries a jurisdiction note, a contract-term flag, and a renewal-risk score.
  • Egress and data-flow map: An egress and data-flow map. We trace every byte from operator input to model output to storage. Country codes mark each hop. We pin PIPEDA principles and ITSG-33 control families to the boundary they cross. Moreover, the map flags every external hop where a Protected B reclassification matters.
  • Phased migration plan: A phased migration plan across four quarterly milestones. Each phase names the workload to land that quarter, the artefact you ship, and the rollback note if the phase fails. Furthermore, each phase lists the procurement leadtime and the named decision-maker on your side.

Prerequisites

  • One AI workload to scope; we work the brief around it end to end.
  • One workload owner who can answer questions live on the call.
  • Any existing topology diagram or vendor list; we work from what you have.
  • Your current data classification posture, even if it is informal today.
  • Honest answers about today's stack, today's gaps, and today's budget.

Who this is for

  • Book the sovereign AI infrastructure Canada brief before signing a hyperscaler contract or a new GPU lease
  • You walk in with one workload
  • Inside scope of the brief

Customize this engagement

Live configurator arrives in milestone 2. For now, mention any custom scope on the kickoff call.

Frequently asked

Do you build, or only brief, for the sovereign AI infrastructure work?

Brief only. Hands-on build sits inside the separate Standing Engagement service or vendor adapter work. The brief gives you the topology, the shortlist, the egress map, and the migration plan; the build phase is a separate engagement you can run with us or with your own team.

Why sovereign for a Canadian AI workload, and what changes if we do not bother?

Sovereign means three things in Canada: a jurisdictional answer that PIPEDA and provincial regulators recognise, an audit-trail your insurer and auditor accept, and Bill C-27 readiness for AIDA when it lands. Without sovereign posture you risk a cross-border data transfer your regulator does not allow. Furthermore, sovereign topology is cheaper to defend in an incident review than a hyperscaler-only stack.

Who actually runs the infrastructure brief, and what is their background?

One Vanwebdev engineer scopes the workload, models the topology, writes the brief. The engineer has shipped Canadian AI infrastructure under PIPEDA and reads ITSG-33 control families for a living. Moreover, the engineer who scopes the brief also writes it; nothing is handed to a junior between the scoping call and the delivered document.

What if our team already has a working topology in place today?

We adapt. If your topology is sound, we audit it against ITSG-33 and PIPEDA boundaries and we cite the gaps. If it is partial, we extend it across the layers it misses. Specifically, the brief always reflects your starting position; we do not propose a green-field rewrite unless your workload truly is green-field.

Will you name specific GPU vendors and specific model hosts, or stay neutral?

We name two or three Canadian-jurisdiction candidates per layer: GPU compute, model hosting, retrieval, vector store, observability, key management. We explain the trade-offs against your throughput envelope and your audit posture. You choose. Vanwebdev does not take vendor referral fees on this brief.

How fast does the brief actually land in our inbox once we book?

The brief lands by Friday the same week the scoping call runs, assuming a Monday or Tuesday call. The engagement covers one follow-up Q&A window of thirty minutes; we book the window inside the brief itself.