From strategy to shipped code. If it's software + AI, I build it — these are the requests I serve most, not the boundaries.
01AI adoption for your business
Where does AI actually pay off in your processes — and where is it hype? Honest assessment, the right stack (cloud vs local, API vs subscription), realistic costs, a rollout plan, and training for your team. No buzzwords, measurable outcomes.
02Custom software & automation
Web apps, backends, integrations, process automation — built AI-augmented, which means fast, and disciplined (strict TDD, 5,800+ tests maintained daily). From a small internal tool to a complete platform.
03Custom assistants & chatbots
Assistants that work on YOUR data — documents, management systems, knowledge bases — with persistent memory and a design that says "I don't know" instead of inventing answers to your customers.
04AI agents for real tasks
Agents that actually execute: paperwork, email triage, document processing, monitoring, reporting — with human supervision where it matters and audit trails everywhere.
05Reliable AI (anti-hallucination)
My specialty. Your AI system invents answers? I fix that: admission gates, blind multi-model validation, provenance, deterministic verifiers. I built one of the few memory layers with an anti-confabulation gate — it ships as Verimem.
06LLM integration & orchestration
Connect AI to your real tools and data (MCP done properly: validation, audit trails, rate limits). 20+ providers integrated — Anthropic, OpenAI, Gemini, local models — with fallback chains, cost caps, no lock-in, no surprise bills.
07AI-assisted security audits
Automated web-security auditing built on my OMNEX platform (50+ phase pipeline, blind validation). Find the problems in your site or web app before somebody else does.
08Something else?
If it involves software and AI, chances are I've built something adjacent. Describe the problem — worst case, I'll tell you honestly that you don't need AI for it. That answer is free.
01Strict TDD
RED → GREEN → regression. A fix without a test that failed before it is not a fix.
02Adversarial gates
My own claims get attacked before shipping: falsification, caller verification, counterexamples.
03Blind validation
Independent models judge without seeing each other — confirmation bias designed out, deterministic verifier last.
04Provenance everywhere
Append-only audit trails, supersession over deletion, facts that carry their source. Not traceable = didn't happen.
Have an AI problem that needs to actually work?
From a one-hour consultation to a complete build: AI adoption, custom software, assistants, automation, reliability engineering, security audits. Everything verifiable is on GitHub; the rest I'll show you live. Remote worldwide, based near Milan.