Stefano Noferi

Founder and CEO/CTO of noze. Tech Entrepreneur · AI Governance & Security Architect

Stefano Noferi

Tech entrepreneur. He works on artificial intelligence, cybersecurity and digital health, with a focus on AI agent governance: agents are becoming autonomous actors inside software systems, but today's architectures were not designed to control them.

He proposed OISG (Open, Intelligent, Secure, Governed), the paradigm he proposed for autonomous AI and adopted by noze, and maintains Admina, the open source framework sponsored by noze that builds governance into AI applications and powers DataGovern, IntelliPA, AgenticHealth and VerdePilot. With DebugABot he does deep tech research on runtime debugging of agents and robots.

In over 25 years he has worked with more than 100 clients and partners and contributed to over 30 European, national and regional research projects with universities, CNR, hospitals and industry. For international organizations, public bodies, universities and media groups he has designed cloud architectures, document systems and web portals, two of which won Il Sole 24 Ore awards.

His background is in hardware, networking and electronics and he has always worked with software and open source. As a radio amateur (IZ5AWZ) he migrated Italy's first radio gateway between AMPRNet and the Internet (IW5DAM) to GNU/Linux at ISTI-CNR in Pisa. He took part in the Linux, Python, Zope and Plone communities and contributed to C.I.R.S., the first Italian consortium of FLOSS companies, to T-OSSLab and to PloneGov Italia.

As a lecturer and coach he has trained over 1,000 professionals in university master's programmes, ITS Academies and companies on AI, cybersecurity, cloud and open source. Founder, mentor and advisor, he supports Italy's startup ecosystem.

Sailing instructor with Lega Navale Italiana and UISP, including inclusive projects. FISAR sommelier and keen cook, interested in music, art, physics and mathematics.

He is naturally curious.

Writes about

Articles

Aleph Neuro: ultrasound brain imaging, fully open source

On 24 June 2026 Aleph Neuro published what it claims is the first 3D image of the vasculature of a living human brain captured with ultrasound through the intact skull, using ultrasound localization microscopy (ULM), releasing the full pipeline and dataset as open source on GitHub. How it works, why it matters for digital health, and our take.

Admina: Open Source framework for AI governance

Admina v0.9.1 released (20 May 2026): Apache 2.0 Open Source framework for AI governance, created by Stefano Noferi and sponsored by noze. Python SDK + Rust proxy, 4 governance domains, ~6µs overhead, 700+ tests.

OISG: noze adopts the paradigm for autonomous AI systems

noze adopts OISG (Open, Intelligent, Secure, Governed), the architectural paradigm proposed by Stefano Noferi for the design and governance of autonomous AI systems. Four interdependent pillars that translate into concrete benefits for products, solutions, partners and clients.

Tutorial: opencode on a remote server via SSH for Linux ops

Using opencode (opencode.ai) inside an SSH session for log triage, package audit and small maintenance tasks on Linux servers. Human-approval flow, no secrets in prompts, preventive backup.

Tutorial: Goose with a read-only MCP server on anonymised FHIR metadata

Block's Goose configured with a custom MCP server exposing read-only metadata of an anonymised FHIR bundle. Local LLM via Ollama, directory isolation, use of test datasets.

Tutorial: Cline in VS Code with manual approval on every action

Cline (formerly Claude Dev) used in VS Code to generate boilerplate and tests with manual review of every diff. Multi-provider configuration, auto-approve disabled, config management for SME teams.

Tutorial: Daniel Miessler's Fabric for structured extraction from documents

Fabric used in a professional firm to apply reusable patterns to tender and contract PDFs. Go binary install, setup with a local provider, extraction of deadlines, amounts, clauses.

Tutorial: OpenHands in Docker for reproducible research experiments

OpenHands (formerly OpenDevin) running in a Docker container with read-only mount on input data. Configurable model endpoint, workspace persistence, network constraints for isolation.

Aider with DeepSeek Coder via OpenRouter: branch refactor

Aider configured with DeepSeek Coder V2 via OpenRouter for a targeted refactor on an isolated branch. Auto-commit, repo map, file-scope constraints, regression tests before merge.

Tutorial: Plandex for plan-then-apply multi-file refactors

Plandex used for coordinated refactors across many files of a legacy app. Plan model, sandboxed changes, per-task rollback, parallel plans without interfering with git.

Tutorial: Continue.dev with on-prem Ollama for NDA-bound code

Continue.dev configured with a local Ollama backend: a code assistant in VS Code/JetBrains with no traffic to the cloud. config.json setup, models for autocomplete and chat, workspace isolation.

Tutorial: Simon Willison's llm CLI for log triage via Unix pipes

Simon Willison's llm CLI used in Unix pipelines to summarise system logs, llm-ollama plugin for local use, preventive redaction of sensitive data before sending to the model.

Tutorial: Open Interpreter for local research dataset exploration

Open Interpreter used in a research environment to explore CSV/Parquet datasets. Install in a virtualenv, --local mode via Ollama, read-only working directory on originals, persistence of generated notebooks.

Need support?Under attack?Service Status
Need support?Under attack?Service Status