Amazon Bedrock pricing and architecture review, SageMaker included

An independent opinion on your Bedrock or SageMaker design before you scale the spend behind it.

Fee: €4,000
Timeline: 2–3 weeks
Format: Async, read-only access or a design export

Before you scale the spend

EU SMBs and scale-ups with an LLM feature in production or on the roadmap
Teams on AWS who want an independent architecture opinion before scaling spend

Relevant credentials

AWS Certified AI Practitioner (early adopter)
AWS Certified Machine Learning Engineer – Associate
AWS Community Builder, AI Engineering

Fixed scope, fixed price

Recommendations are vendor-independent and based on public AWS documentation, your own data and my general experience. No third-party or employer information is used.

GenAI architecture assessment

€4,000

2–3 weeks

Scope: Your current or proposed design: model and provider choice, RAG and data flow, guardrails, observability, cost model, and security and data residency in EU regions.
You get: A target-architecture diagram, a 10–15 page findings report with a prioritised backlog, a 12-month cost estimate, and a two-hour evening walkthrough.

Add-on: AI Act alignment

€1,000

+1 week

Scope: Maps the assessed system to its AI Act risk tier and the deployer duties that follow.
You get: A three-page annex to the main report.

Excl. VAT. EU businesses with a valid VAT number are invoiced reverse-charge. Invoiced on delivery, 14-day payment terms, after a signed one-page scope.

Not sure which AI Act tier your system falls into? Start with the free AI Act checklist.

Three steps, asynchronous

Step 1 of 3
Scoping call
A 45-minute evening call, then read-only access to your account or an architecture export.
Step 2 of 3
Async assessment
Two to three weeks of work on my own time, with written questions as they come up.
Step 3 of 3
Report and walkthrough
You get the report, then we walk through it together in one evening session.

What I don't take on

I do not work with my employer, with any company in its group, or with their competitors, suppliers and business partners.
I use no employer information, tools, equipment or materials, and I work only outside my working hours. All benchmarks come from public sources and your own data.

Request a scoping call

Booking form not loading? Open it in a new tab or email me directly.

Assessment questions

What if the workload is not built yet?

That is the cheaper time to do this. I review the proposed design and the cost model before you commit to a model, a vector store or a region.

Do you recommend specific models?

Yes, with reasons and a cost per thousand requests. I have no reseller relationship with any provider, so the recommendation is based on your workload rather than a margin.

How do you estimate 12-month cost?

From your expected request volume and token profile, priced against published AWS rates, with the sensitivity ranges that matter. You get the spreadsheet, not just a number.