A fixed price, in writing
Scope and fee are agreed before work starts. No open-ended hourly billing.
Services and fees
Scoped work with a written price
For engineering leads and founders who need a production system, not a strategy deck. Each engagement has a defined deliverable. You get a single price after a short scoping call.
01 · Typical starting point
Production Readiness Audit
€4,500
Fixed fee · 5 working days
For teams that want to use AI and are not sure what to build, or that already have a prototype and need a clear view before spending more.
What happens
- Day 1: kickoff to map goals, systems, data, and constraints
- Days 2–4: review of data sources, infrastructure, any existing prototype, and workflows
- Day 5: written report and a 60-minute readout
What you get
- Ranked use cases with expected value, effort, and risk
- Build versus buy, with named vendors where relevant
- Data readiness and security gaps
- An architecture sketch for the highest-priority use case
- A 90-day plan, including work that should not be done
Credit
If we move to architecture or a build within 30 days, the €4,500 is credited in full.
02
AI System Architecture
€12k–€25k
2–4 weeks · fixed after scoping
For teams that know what they want to build and need it designed so it can pass security review and be operated by the people who already run your software.
What is designed
- Approach: RAG, agents, classical ML, or ordinary software where that is enough
- Data flow, ingestion, and permissions
- Model selection and routing, with fallbacks
- Evaluation: golden sets, metrics, and acceptance thresholds
- Observability, cost budgets, and failure modes
- Security notes specific to LLM systems
What you get
- Architecture document and diagrams
- A starter evaluation set and baseline measurements
- A sequenced, costed build backlog
- A handover session with your team
03
Pilot to Production
€18k–€40k
4–8 weeks · one workflow
The first production version in your repository: one workflow, one integration, one success metric. Kept narrow so it actually ships.
Typical builds
- Internal retrieval over documents, tickets, or a graph, with citations
- Metadata extraction, classification, and multilingual pipelines
- Tool-using agents behind an API, with search and fallbacks
- Scoring, forecasting, or image models served as an API
What you get
- Working, deployed code in your repository
- An evaluation harness
- Tracing, dashboards, alerts, and fallbacks
- A runbook and handover against criteria you signed off
04
Fractional AI Architect
€4,500/mo
1 day a week · monthly
For teams with AI in production and no senior owner for the next set of decisions. Architecture support without a full-time hire. A second day a week is €4,500 more per month.
Included
- Architecture reviews for new features
- Review of AI-related pull requests and designs
- Model upgrade testing and rollout
- Vendor evaluation and build versus buy
- Monthly notes on quality, latency, and cost
- Priority support when something breaks
Terms
Three-month minimum, then month to month. 30 days' notice either way. The fee above is one day a week. A second day is agreed separately, at €4,500 more per month.
Working together
What to expect
A written update each week
What shipped, what the numbers say, and what is next.
Your environment
Your repository, your cloud, your tools. Nothing locked into a platform of mine.
A clean handover
Documentation, a runbook, and a walkthrough so your team can run the system.
Questions
Common questions
Do you also do classical machine learning?
Yes, when it is the right tool. Scoring, forecasting, ranking, and image models are often cheaper and more reliable than a language model. That work sits inside the same architecture and build engagements.
Why are fees published?
So you can tell whether this is in budget before we speak, and so the page can be forwarded internally. Ranges become one fixed price after a short scoping call.
Do I have to start with the audit?
No. If the use case is already clear, we can go to architecture or a build. The audit is there for teams that are not at that point yet.
Who owns the code?
You do. Work is done in your repository and your cloud account, documented so your team can run it without me.
Will you recommend not building something?
Yes. A slow process is often a process problem. Putting a model on top of it makes that worse. Sometimes the right answer is ordinary software, a vendor product, or nothing.
Where are you based?
Malta. I work with teams all around the globe.
Not sure which engagement fits?
Book a 30-minute call. I will say which one makes sense, or that none of them do.