01 · Artificial intelligence
Integrating AI and LLM agents into your applications, all the way to production
I'm a senior developer specialising in backend and infrastructure. I integrate language models (OpenAI, Anthropic, Gemini, Mistral, or open-weight models hosted on your own servers) into existing applications and team workflows. My goal: a useful, measurable feature whose cost and reliability are under control.
AI in production, not in a demo.
What I do
- AI features in an existing product
- Writing assistants, content generation, summaries, classification, smart search, wired into your application and API.
- AI agents
- Agents that act through tools (APIs, databases, code repositories, tickets), with guardrails, logs and human approval where needed.
- Search across your documents (RAG)
- A queryable knowledge base, embeddings, PostgreSQL + pgvector, source citations.
- Document extraction (OCR, vision)
- Invoices, forms, supporting documents: structured data from PDFs and images.
- Process automation
- Repetitive tasks, triage, routing, reporting.
- Local models and privacy
- Inference on your own servers when data must not leave them: model choice, quantisation, GPU sizing.
- Cost and latency
- The right model for each task, caching, queued processing, cost tracking.
Examples
-
An agent that turns a production error into a merge request
From a Sentry alert, it reproduces the context in an isolated Docker sandbox, proposes a fix, runs the tests and answers developers' questions on the merge request.
AI agents for a development team — Read the case study -
A code review agent
Bugs, security, performance, readability, release notes and DORA metrics.
AI agents for a development team — Read the case study -
A monitoring tool with an investigation agent
Response times, slow queries, service-level objectives (SLOs), and an agent that analyses anomalies and suggests leads.
AI agents for a development team — Read the case study -
A backend migration driven by AI agents
Personal project: Laravel to NestJS, domain by domain, with automated checks, adversarial review and the old code as a test oracle.
Galactic Reigns — Read the case study -
Assistants built into HR software
Email drafting, survey creation, event organisation, and a conversational agent backed by a knowledge base.
Approach
-
01
Scoping (2 to 5 days)
Use cases, available data, GDPR constraints, API or local model, measurable success criteria.
-
02
Prototype on your data
We check quality before industrialising.
-
03
Production
Integration, tests, monitoring, cost tracking, documentation.
-
04
Handover
Your team can run and evolve the feature.
Technologies
- PHP / Laravel
- TypeScript / NestJS
- Python / FastAPI
- Vue 3
- PostgreSQL / pgvector
- Docker
- AWS
- OpenAI
- Anthropic
- Gemini
- Mistral
- llama.cpp
- Open-weight models
Rate
From €900 per day, excl. VAT
or a fixed price after scoping.
FAQ
Can you add an LLM to an existing Laravel application?
Yes. I connect the model to your code and data, with queued processing, caching and cost tracking.
API or local model: how do I choose?
APIs (OpenAI, Anthropic, Gemini, Mistral) are simpler and often more capable. A local model makes sense when data must stay on your servers or when volume makes the API too expensive. I help you decide, with numbers.
What about GDPR?
We decide which data is sent to the model, where it's hosted (EU or on-premise), retention and access, from the scoping phase.
How much does an AI agent cost?
Development depends on scope; running costs (inference) are measured from the prototype onwards, so you can decide with the facts.
Got a project in mind?
Describe what you need in a few lines: I'll get back to you with a first analysis and next steps.