AI systems built to run in production, not just in a demo.
RAG assistants over your documents, LLM features inside your products and AI agents that act on real systems, deployed on AWS or GCP with the discipline of production infrastructure: CI/CD, monitoring, evaluation of answer quality and limits on model cost.
What I build
RAG assistant over your documents
Ingestion, chunking, embeddings and retrieval, with answers that cite their sources.
LLM features in your product
Model APIs wired into existing products and internal tools: prompts, tool use, structured output, latency and cost control.
AI agents
Agents that call your APIs and tools, with scoped permissions, an audit trail and human approval before irreversible actions.
AI applications end to end
From prototype to production: the Python backend and APIs, authentication and cloud infrastructure behind the AI feature.
Evaluation and guardrails
An evaluation set for every project, so answer quality is measured before and after each change, not assumed.
Running it in production
Kubernetes or serverless, Terraform, CI/CD, secrets management, monitoring and spend dashboards.
Proof you can check
Built end to end: retrieval over the CV, FAQ and articles; a retrieval and answer test suite that must pass before every deploy; rate limits, a daily cost cap, no visitor IPs stored.
Delivered AI engineering projects: LLM integrations and RAG pipelines.
My UK consultancy: cloud infrastructure, automation and AI engineering for clients across Western Europe.
Kubernetes, Terraform, CI/CD and observability on AWS and GCP: the part that keeps an AI feature up once it is live.
Not on offer: general web or mobile apps without an AI or infrastructure component.
Describe the problem
Current setup, what is not working or what you want to build, and any deadline. Use the form below.
Short scoping call
I reply personally and propose a short call to scope the work, before any commitment.
Contract and delivery
B2B contract through Digital Cloud OPS Ltd (UK). Delivered remotely from Bucharest, with full working-day overlap with UK, Irish and Central European teams.
What would you like to automate or answer?
A few lines about the documents, systems or users involved. I reply within 24 hours.