Anthony A.

I'm a Senior Software & AI Engineer, specialising in privacy-preserving LLM deployments and regulatory AI — the area where most enterprise GenAI projects in Germany either succeed or quietly fail.

Over 8+ years I've shipped production systems at Nokia (distributed microservices, 4 years), built a RAG privacy platform for enterprise clients handling sensitive data, led technical research for the European Banking supervisory AI framework — the most stringent financial-regulatory environment in Europe — and currently architect a self-hosted LLM evaluation platform for Heidelberg University with 100% data sovereignty.

Where I add the most value

— Self-hosted / sovereign LLM deployments. Local inference with Ollama and vLLM. Llama, Mistral, and GPT-class models running on your hardware, your VPC, your governance. Zero data leaves your perimeter.

— RAG systems for regulated industries. Document ingestion, chunking, embedding, vector retrieval, LLM evaluation agents, structured output. Production architecture from day one — not a notebook demo. PII masking and sovereign data handling built in.

— GDPR-compliant GenAI. Context-injection pipelines designed for compliance-sensitive workflows in banking, government, HR, and healthcare.

— Full-stack delivery. FastAPI backends, React dashboards, PostgreSQL, Docker, Kubernetes. I ship the whole feature, not just the model.

How I work

Bias toward shipping testable increments over long demos. Strong written communication — architecture documents non-technical stakeholders can actually read. Remote-first across DACH and EU, onsite possible.

Engagement shapes

— De-risking sprint (1–2 weeks): scope and prototype a GenAI feature before you commit headcount.

— Feature build (1–3 months): RAG system, AI assistant, document evaluation pipeline, internal copilot. Architected for production.

— Fractional AI engineer (3+ months): embedded with your team to add GenAI capabilities to an existing product without breaking what works.