Skip to content
QADtechLab

Services

Three layers. Six things we will actually take money for.

The thesis covers software, infrastructure and intelligence. What we sell is deliberately narrower than what we publish — a firm claiming all three layers at launch reads as one that does none of them well.

Layer 1

Software

Renting: SaaS subscriptions — per seat, per month, forever, rising annually. Owning: Custom applications we build; you hold the repository.

  • Launch Sprint4 weeks · Fixed price

    One core workflow, live in four weeks.

  • Tool Replacement6 weeks · Fixed price

    One subscription, replaced by software you own.

  • Platform Replacement10 weeks to the first system · Milestones

    First system in ten weeks, then a costed roadmap.

  • Own & OperateAnnual (AMC) or monthly (operated) · Cancel anytime · exit handover included

    We keep it working, or we run it. You keep owning it either way.

Layer 2

Infrastructure

Renting: AWS, Azure, GCP — metered compute, storage and egress. Owning: Hardware in a colo or your own rack.

  • Cloud Exit Assessment2–3 weeks · Fixed fee

    Whether to leave the cloud — including the answer “don’t”.

  • Cloud Cost Optimisation2–3 weeks · Share of savings

    Stop overpaying for infrastructure you are going to keep.

    Terms are settled; the detail page is still being written. Ask and you get the same answers by email.

Published as direction — not sold

  • On-Prem Migration3–9 months · Milestones

    Moving workloads out of metered cloud and onto hardware you own, in a colo or your own rack.

  • Managed On-PremOngoing · Retainer

    Running the hardware after the move — the answer to "who owns 3am?", which is the question that decides most repatriation deals.

These are on the roadmap, not the price list. What has to be true before we sell them.

Layer 3

Intelligence

Renting: Per-token API calls, with your data leaving the perimeter. Owning: Open-weights models on your own GPUs.

Published as direction — not sold

  • AI Readiness Review2 weeks · Fixed fee

    Which of your workloads should run on your own models, which should stay on an API, and which should not use a model at all.

  • Private AI Stack6–12 weeks · Milestones

    Open-weights models on your own GPUs, for the narrow high-volume work where that genuinely beats an API — classification, extraction, tagging, routing, embeddings.

These are on the roadmap, not the price list. What has to be true before we sell them.

The rule

The teardown picks the engagement for you.

Four minutes, no email, computed in your browser. It ranks your own stack by payback and routes each verdict to the service that delivers it — including the tools it tells you to keep.