
I measure how decisions get made: moderation quality, policy enforcement, and the AI systems that now take part in them. Seven years, Dublin, at Meta and TikTok. The thread through all of it: I build and measure the systems that decide, and I keep a person in charge of the decision.
Integrity operations and root-cause work. Co-developed a reusable root-cause analysis approach for integrity quality issues; the quality-measurement dashboard prototype was accepted as a basis for further development.
Evaluated LLM-judge policy decisions on fraud, scam and phishing content and contributed to the evaluation rubric.
Search relevance and quality evaluation, calibration, error analysis; led a team of eleven.
Redesigned the child-safety escalation path with a risk matrix, decision tree and SOPs: SLA compliance from 76% to 97%, processing time from 48h to 33h.
A vendor project, not direct Google employment.
Customer support, marketing and project coordination, in Ireland, Portugal and Australia.
Every number I use has a source I can show. My profile is a register of facts with provenance; when a number is missing I say so, and when a system is uncertain I would rather it showed the uncertainty than hid it.
When I built JIOS I held out 65 of my own past decisions the model had never seen and tested four versions of what it reads about me. One scored 92.5%; the others lower. Only the measured best went live, and the measurement stays on the page.
Rules filter, the model reads and explains, a person makes the call. I design it that way at work and in what I build: nothing is applied for, enforced or scored without a human who owns the outcome.
I have lived and worked in Italy, Portugal, Australia and Ireland. I studied linguistic mediation in Sassari and communication for international relations at IULM in Milan; I work in Italian and English. I build tools for the problems in front of me: JIOS, a job-intelligence system; Ramona, a personal assistant with memory; a safety and quality gate for LLM output. I like problems where the policy is clear and the data is not.
A Trust & Safety, integrity or AI-quality team in Dublin, hybrid or on-site, where the work is measuring and improving how decisions get made: moderation quality, policy enforcement, LLM evaluation, or program management inside such a team. Mid-level, hands-on, with room to own a quality programme. Not: sales, pure engineering, or a role where the model decides and nobody checks.
Email and LinkedIn both reach me. CV on request.