Available for new projects · Remote · Milan, Italy
Trusted by teams at H-Farm, GB Viaggi, Pellemoda, and more · GDPR & EU AI Act–ready.
Live trace of a production agent run: routing, retrieval, GDPR guardrails, actions, and LangSmith observability.
Pellemoda S.r.l. — An autonomous inventory agent for Pellemoda — demand forecasting and supplier coordination on autopilot
H-Farm College — How H-Farm College automated an admissions pipeline handling 2,000+ applications a year
GB Viaggi — How GB Viaggi automated 3,000+ bookings a day and unlocked €500K+ a year
More in the case studies.
“The automation handles thousands of operations daily without manual work. It completely changed how we operate.”
See the code
MemGuard →Open-source agent-memory validation — continuously verifies that facts stored in agent memory (Mem0, Zep, Letta, LangMem) are still true. See exactly how I write production agent code.
The lowest-risk start is a Production-Readiness Audit (€1,500, one week — credited if we continue). Most clients then run a fixed-scope Sprint (€8,000, 2–4 weeks) and graduate to a Build for the full production system. See services & pricing or get a tailored figure with the cost calculator.
An AI agent is a system that uses an LLM to reason over a goal, decide which steps and tools to use, and act — not just answer a single prompt. In practice that means things like reading documents, calling your APIs, updating a CRM, or escalating to a human when it's unsure.
The lowest-risk start is a €1,500 Production-Readiness Audit (one week, credited if we continue). A fixed-scope Sprint (a working agent in 2–4 weeks) starts at €8,000; a full production system runs €18,000–€35,000, plus modest monthly running costs. Try the AI Agent Cost Calculator for a tailored figure, or book a free call.
2–4 weeks for a demoable Sprint, 6–10 weeks for a full production Build. If you already have a product and just need AI features added, that's typically a Sprint.
If a SaaS tool already does 90% of what you need, buy it. Build custom when the agent has to work with your data, your workflows, or your compliance constraints — that's where a generic tool falls down and a tailored agent pays for itself. I'll give you an honest build-vs-buy answer on the call.
Yes — retrofitting agents, RAG, and search into an existing stack is one of the most common engagements. I work remotely with startups and SMBs across the EU and English-speaking markets, in English or Italian.
Evaluation sets, LangSmith observability, guardrails and PII filtering, human-in-the-loop on risky actions, and full audit trails. The goal is a system your team — and your legal team — can actually depend on.
100%, from day one. No lock-in, no proprietary black boxes; Builds include handover documentation.
Book a free 30-minute call — tell me what you're building and I'll tell you if I can help and what it would take.
Not ready? Get my AI-agent build checklist + occasional notes.