AI integration and AI agent consulting - for real, no buzzwords
Custom AI integration, AI agents, AI automation and AI consulting. A senior developer with own AI products (MindBeamer, ChatKraft) - I build things that work, not presentations.
What I build with AI
Concrete integrations, no PowerPoint decks. Agents, automation, chatbots - all running in production.
Custom AI agents
Agents that perform real tasks: query databases, manipulate files, send emails, interact with external APIs. Using Claude, GPT-4, Gemini or local models depending on what fits.
AI automation
Intelligent workflows that process documents, classify tickets, generate reports, moderate content. Connected to your existing stack - Slack, Notion, databases, webhooks.
Professional chatbots
Chatbots with your knowledge (RAG), defined personality and no hallucinations. Integrated into your website, WhatsApp, Telegram or internal panel. No templates, all custom-built.
AI integration in existing apps
I add AI capabilities to your current applications: semantic search, automatic summaries, text analysis, image classification. Without rewriting your stack.
Own AI products - not just presentations
The difference between an agency that sells hype and a developer who builds: real products in production with real users.
Palimpalem - Onboarding with AI chat
Website builder with AI chat onboarding: the assistant discusses topic, colors and writing style with the user, uploaded images are analyzed and the best ones placed in the most prominent spots. Right after the chat, a complete website is ready to edit. Huge time saver. 1.4M+ sites since 2004. In production.
palimpalem.comklaro.legal - AI for contract review
SaaS platform that analyzes contracts with AI: detects problematic clauses, compares with industry standards, and delivers an actionable report in minutes. GDPR-compliant with European model. In production.
klaro.legalMindBeamer.io - AI content automation
SaaS platform that automates social-media content creation with AI. Multi-agent, multi-platform. Currently in active development - architecture complete, first pilot customers.
mindbeamer.ioHow we work together
Four steps, no vaporware. From first idea to production.
Concept
We talk about the real problem. AI gets discarded if it doesn't add value - sometimes a well-designed database is better.
Prototype
In 1-2 weeks you have a working MVP with real data. So you decide based on something tangible, not slides.
Production
If the prototype convinces: infrastructure, monitoring, rate limiting, cost control.
Iteration
Continuous improvements based on real usage. Models evolve fast - we adapt when something better appears.
Curious what building AI agents in production actually looks like? I write about it on the blog: Autonomous agents are the easy part
Frequently asked questions
Which AI model do you use - OpenAI, Anthropic, Google, local?
Whichever fits best. Claude (Anthropic) for reasoning and code, GPT-4 for general tasks, Gemini for multimodal, and local models (Llama, Mistral) when privacy or cost requirements demand it. I explain the trade-off before choosing.
How much does an AI integration project cost?
Depends on scope, and honestly, AI has just shifted what things cost and how long they take. So I won't quote numbers here that stop being true next month. Describe your project and you get a binding fixed-price quote after the concept phase.
What if a better model comes out tomorrow?
That's why I build everything with provider abstraction. Switching from Claude to GPT-4 or vice versa is a configuration change, not a rewrite. Your architecture stays yours, models are interchangeable.
Do you handle infrastructure and API costs?
Yes. Rate limiting, caching, retry logic, cost monitoring - all included in production. I save you from surprise €5,000 bills due to infinite loops or API abuse.
What about GDPR and sensitive data?
For sensitive data I use models with European DPA or local on-premise models. We clarify this in the concept phase - GDPR is not an add-on, it's an architectural constraint.
Real clients, real words
A few honest words from people I've worked with over the years.
1 / 3
Want to be next? A no-commitment 15-minute chat on WhatsApp, that's how most of these stories started.
Let's talkReady for real AI integration?
Message me on WhatsApp with your idea. In the first conversation we discuss whether AI adds value, which model fits, and what to expect in costs and timelines. Honest from the first message.
Message me on WhatsAppPrefer to write in your own time?