Applied AI / LLM Engineer | Claude in production, LLM evaluation, EU data residency | Python

I build LLM systems that run in production and pass a compliance review.

My own product, Esmond, is an evaluative AI that assesses academic work before submission. It runs unattended from request to delivered report, in three languages, for paying customers: LLM orchestration on Anthropic (Claude) and OpenAI models, multi-step pipelines, prompt architecture and prompt caching, evaluation harnesses, evidence verification, cost monitoring per run. API in Python (FastAPI), PostgreSQL, payment flow, infrastructure and monitoring: built and operated single-handedly. I have also built integrations for AWS Bedrock, Google Vertex AI and OpenRouter.

Compliance is part of the design, not a layer on top: AI Act qualification (Annex III), GDPR retention, processing agreements and transparency, EU data residency, zero data retention. Behind that sit ten years as director of an investment vehicle registered with the FSMA, the Belgian financial regulator: periodic reporting, customer due diligence under anti-money-laundering law, and the platforms that produce the reporting.

Data and integration work: PostgreSQL, SQL, data modelling, ETL from point-of-sale, ERP and accounting systems, demand forecasting with a measured error margin. One Odoo project through the XML-RPC API (point of sale, sales, inventory, manufacturing, accounting), developed on pre-production and rolled out to production.

Stack: Python (FastAPI), JavaScript and React, SQL, Astro, Docker, Railway, Linux, Stripe, Supabase, automated testing, Playwright, PyMuPDF.

Master in Germanic Languages and sixteen years of adult education: I explain technical work to people outside it, in Dutch, English and French. Basic German.

Freelance through Ramirez BV (BE 1039.976.194), Ghent, Belgium. Fully remote, CET. Time and materials per day or a fixed price on a defined scope; also as a subcontractor for agencies. A defined scope, a working system, documentation and handover.