I build the agentic foundation a team makes data-driven decisions on, guide organisations through AI adoption, and run the marketing and growth operations — tooling, automation, measurement — turning both into daily practice. I work alongside teams through training programmes and continuing strategic support, fractional where that fits. Neither an agency nor a software house.
Most already know roughly what they should be doing. What stops them is that the numbers do not speak to each other, the tools do not communicate, and half the week goes to assembling reports by hand and reading them — time the team does not get back.
My work is informed by strategy but starts at the foundation: deciding what to measure and what to model, getting the data where it is needed without anyone copying it, building the models that turn those numbers into a decision, and — where it helps — bringing AI into the workflow to the point where it genuinely removes repetitive work.
I can build and prototype all of this, drawing on an extensive personal codebase. For the final development I work alongside your team, or with a partner I can bring in.
What to measure and what to model, before the question of which tool to buy. Statistical modelling, data mining and machine learning on the data you already hold: forecasting, segmentation, experiments, and an intuitive reading of the result.
Driving adoption across the organisation - not delivering a pilot and leaving. The use cases that actually pay, the real cost of running them, who has to change how they work, what the AI Act obliges you to document. Certified in AI & Law and AI Governance.
The plumbing that makes a team autonomous: a tracking plan people follow, events and pipelines into one place, CRM and lifecycle flows that fire unsupervised, dashboards with a reporting cadence. One agreed definition of a conversion, and no more assembling Monday's numbers by hand.
The hard calls, without a box on the org chart: is this feasible with the data and people you have, build or buy, which vendor, who to hire first. Where an existing tool already solves the problem, I say so rather than open a building site.
Marketing and growth. Education. Operations and internal knowledge. Scientific and research data — signal processing, machine listening and experimental datasets are where I started, at IRCAM, and remain fully in scope.
Training that ends up inside the processes, measurement that holds the various platforms together, infrastructure that absorbs growth. Three recent engagements, told in full.
EU policy intelligence platform. Knowledge representation, information extraction, ingestion and retrieval over legislative corpora that change weekly, semantic search, and assisted drafting of documents.
Italian online tutoring. Omnichannel measurement framework across Google Ads, Meta, LinkedIn Ads and TikTok, proprietary tooling that builds editorial plans against the stages of the sales funnel, and an AI tutor covering the hours human tutors cannot.
Server and client for near-realtime audio transcription: WebSocket streaming, Huggingface voice activity detection and Whisper (faster-whisper) for recognition. Modular design for swapping VAD and ASR components, configurable chunking strategies, multilingual support.
I set up an omnichannel, multitouch measurement framework, with shared definitions and data flowing into a single model, instead of a thousand dashboards telling different stories.
Proprietary software produces editorial plans matched to what the sales funnel actually needs built: every piece of content and every euro answers to a specific stage, and measurement comes back on the same scheme.
I speak at public conferences, keynotes and panels, and teach university courses and in-house workshops, in English, Italian or French. The recurring subjects are AI adoption inside organisations, data strategy, and data-driven marketing — treated as operational problems with real examples, not as industry overviews.
Instrument, connect, automate. At the end the team has one agreed set of numbers, the reports arrive on their own, and the first two or three automations are live and documented.
Working inside the team on a standing basis: keeping the stack honest, extending the automations, running the measurement cadence, and driving AI into workflows as the team becomes ready for it.
Lighter-touch work for founders and leadership: regular sessions, decisions reviewed, technical claims checked, feasibility read before spending. Retainer or, early-stage, partly equity.
For any project worth building. If the idea is interesting and there is room to do it properly, we find the form: collaboration between peers, research, a shared prototype, an open-source contribution. Tell me what you have in mind.
It is usually the fastest way in. The first conversation covers what the team measures today, where the manual work goes, and what could be automated or answered by AI within a month. If what you need is an agency or a development shop, I will say so.