Work together
AI & automation in production
I build AI and automation that runs in production — in your systems, on your data, with its accuracy measured before anyone relies on it.
Most AI projects stall between a promising demo and something a team can depend on. Integration, evaluation, cost, failure handling, compliance — that gap is the part I close.
What I can build for you
An AI assistant on your company knowledge
Answers from your documents, procedures and past tickets — with the source cited, and an honest “not in the documents” instead of a made-up answer.
For internal knowledge bases, customer support, onboarding
Document processing
Invoices, contracts, forms and emails turned into structured data in your ERP, CRM or database. Uncertain cases go to a person instead of straight through.
For finance, back office, operations
Request triage and routing
Incoming emails, tickets and forms classified, prioritised and sent to the right person — with a draft reply already waiting.
For support desks, shared inboxes, service teams
Process automation
Repetitive multi-step work across your tools — reports, data entry, follow-ups, hand-offs. Plain rules where rules are enough; a language model only where it earns its cost.
For any team losing hours a week to copy-paste work
AI features inside your product
LLM capabilities added to your existing backend, with guardrails, cost limits and a fallback for when the model gets it wrong.
For software companies and product teams
From prototype to production
Your pilot works in a demo. I make it dependable: evaluation, monitoring, cost per request, timeouts, degraded mode, an audit trail.
For teams whose AI pilot is stuck
EU AI Act readiness, from the engineering side
Which of your systems fall under the Act and in what role, and what to fix first — logging, human oversight, documentation. Alongside your lawyers, not instead of them.
For companies building or deploying AI in the EU
How it works
- 01A 30-minute call. You describe the process. I tell you straight whether AI or automation is worth it — sometimes the answer is a simpler fix.
- 02Scoping. I map the process and the data, agree one success metric with you, and send a proposal.
- 03Pilot. A working version on your real data, measured against that metric. You see the numbers before you commit to more.
- 04Production. Deployed with monitoring and documentation, and kept running with my support.
What you get
- Working software, not a slide deck — in your infrastructure or the cloud of your choice.
- An evaluation set built from your own data, so accuracy is measured, not promised.
- A known cost per request before go-live.
- A person in the loop wherever a mistake is expensive.
- No lock-in to one AI vendor — OpenAI, Anthropic, open-source or EU-hosted models.
A good fit
- A process that eats hours every week
- Knowledge spread across documents nobody can find
- An AI pilot that never reached production
- Data that has to stay in the EU
Not a fit
- “We need AI” with no process behind it
- A chatbot as a marketing feature
- A demo for next week’s board meeting
Have a process in mind?
Book a call, or send one paragraph: what happens today, how often, and who does it.