AI engineering for PHP systems · Remote & on site · Germany & Europe
Integrate AI into your existing PHP software — no rewrite, no demo project
Your system has run for years and carries the business. I build AI features where they deliver measurable value: into your workflows, your data, your code. As a senior PHP developer with a Master's degree in Intelligent Systems, I connect both sides — the AI concept and the engineering reality.
Does this sound familiar?
Management wants “something with AI”. The business side has concrete pain points: reports someone evaluates by hand every week. Incoming emails and documents that are sorted and keyed in manually. Support requests that need the same answers over and over.
The software where all of this happens is a grown PHP application. It was not built for AI, but it holds the data and the processes that matter. A rewrite is out of the question. An isolated AI demo next to the system achieves nothing if it never reaches the daily routine.
This is where I come in: AI as a maintainable feature inside your existing software — not as a foreign body next to it.
What the integration delivers
Less manual work
Recurring evaluations, classifications and replies run automatically inside your existing process. Your staff review results instead of producing them.
Decisions based on your data
AI-assisted dashboards and explanations of reports, straight from the data your system already holds. No additional data platform required.
Production-ready, not a prototype
Error handling, cost control for LLM APIs, tests, monitoring and deployment are part of the job from day one. The result is a feature your team can keep running.
Typical use cases
Analysis & insights
AI-assisted analysis and explanation of reports, AI-supported dashboards, automated evaluation of documents, emails and data.
LLM integration
Connecting LLM APIs to PHP applications — Symfony, Laminas, Mezzio, Laravel or custom architectures. Backend logic for AI-powered SaaS products, prompt-driven internal tools for operations, support and content.
Process automation
Automation of recurring manual processes with APIs, queues and background jobs — with AI only where it makes the difference.
AI tooling for your development team
Introducing coding agents such as Claude Code to PHP teams: rules, hooks, review process and limits. My Claude Code cheat sheet shows what that looks like in practice.
Models to match the need
Claude, ChatGPT and Codex where cloud models fit. Qwen, DeepSeek and other open models via Ollama on your own infrastructure where data must not leave the building. Kling for video generation.
Modernise first, then AI
If the codebase cannot be changed safely, legacy PHP modernisation is the first step. Both come from one source.
This has been done before
Nextcloud: recommender engine for the Nextcloud app ecosystem based on collaborative filtering and content-based recommendation — machine learning in production inside a PHP platform.
DMARCFlow (own SaaS product): automated analysis of DMARC reports with reporting and forecasting logic, built with PHP 8, Laminas, Mezzio and React.
TEVEO: a new e-commerce backend from scratch, with Claude Code as the central development tool and markedly faster delivery as a result — AI in the development process, not as a product feature.
How to engage
AI opportunity assessment (fixed price)
One to two weeks. I look at your system and your processes and name the two or three places where AI brings measurable value at reasonable cost. Result: a prioritised list with effort estimates.
Integration of one AI feature (fixed scope)
One clearly defined use case, implemented production-ready inside your existing software — including tests, monitoring and handover to your team.
Ongoing support (retainer)
Further development, operation and cost control of the AI features, plus sparring for your development team.
Rates are discussed in the first call, once the scope is clear.
Frequently asked questions
Does our system need to be rebuilt for this?
No. The integration happens inside the existing application. The prerequisite is a codebase that can be changed safely. If that is not the case, modernisation is the first step.
Which AI providers are an option?
That depends on data protection, cost and use case. I work with Claude, ChatGPT and Codex as well as with locally hosted models (Qwen, DeepSeek via Ollama) and advise provider-independently in the first call.
How do you handle GDPR and sensitive data?
The biggest lever is running local models on your own infrastructure: the data never leaves your premises. Where cloud APIs make sense, only the data the use case really needs goes to the provider, and only under its data processing agreement. We settle which option fits before implementation starts.
Do you work remotely?
Remote and on site in Germany and across Europe.
Already built with AI — and unsure whether it holds?
Your team shipped fast with AI tools and you want to know whether the result survives production? There is a dedicated offer for that.
Review, security check and cleanup of AI-generated code — before your customers find the problems.
See the offer →Let's talk about your use case
A first call, 30 minutes, no obligation. You describe the process, I tell you whether and how AI can be integrated there in a meaningful way.