Twenty years in clinical trials and systems management
I’ve worked on research trials and run pharmaceutical trials across Europe. Then I set up and owned the systems those trials rely on.

I started on research trials as a technician in 2003 and became a clinical research associate (CRA) in 2007. By 2014 I was managing pharmaceutical trials at country and European level, through to close-out. Then I moved to the systems side. I implemented a clinical trial management system (CTMS), evaluated eCOA, IVRS and EDC, and in time owned these systems outright. For years I did both jobs: I specified the tools and lived with what they did on real trials.
Results are only trusted when the work behind them is validated and compliant, and I hold my work to that standard. I define the scope, record what happens and keep the proof. AI needs the same habits if people are going to trust it.
This site explains how I work, and it is the quickest way to reach me.
The record
What I’m working on
I have four projects running, and epiClavis matters most to me. It brings validation discipline to software that AI writes. Every change has a reason, decisions are recorded and the evidence is kept as the work happens. I use it on all my projects, epiClavis included.
Open a project for the details ↗
PRJ-01epiClavis: design control for AI-written softwareAn engine that treats an AI-assisted codebase the way pharma treats a validated system.FlagshipPRJ-02epiEspañol: Spanish that sounds nativeA translation tool for English speakers living in Spain, with a teaching layer built on a collection of sentences graded by CEFR level.Not yet livePRJ-03Screening, KPI and QTL dashboardsRecruitment, screening, KPIs and QTLs for a fictional CRO, planned and monitored against published evidence.LiveSVC-01epiCursor: a web studio for local businessesWebsites in Spanish, Catalan and English for small businesses on the Catalan coast, shown on one site that restyles itself for each trade.In productionWhere AI fits in clinical work
AI is a new risk for pharma, and pharma already knows how to assess risk. The useful question is which tasks AI should do, and what risk each one carries. I’d keep it away from the regulated core and use it for the routine work around it.
Less time on routine work
The old rule says you can pick only two of speed, cost and quality. That trade-off is looser than it was. AI can now take on the routine, documented tasks that eat up time, and the quality standard stays where it was.
Each task has a scope, a record and a reviewer
AI is often pictured as chaotic, with data going everywhere. In practice each task is scoped, every change is recorded and a person reviews and approves the result. That is a controlled change, and pharma has managed those for 40 years.
Built to the standard it will be held to
Low risk is still risk. A tool built quickly still needs records, traceability and an explanation that holds up months later. epiClavis is built to provide that discipline. It is new and not yet proven on a live audited system.
The spreadsheet nobody else understands
I don't mean your EDC or your eTMF. I mean the work around them: visit trackers, quality control (QC) timelines and spotting outliers worth a closer look. I have built tools like these before in Excel and VBA, because that was what I had. The risk is low, the work takes real time and these tools often keep a study running.
What the regulations ask for, and how agents can produce the evidence as they work. Read the full piece ↗
Let’s talk
If you’re implementing or running a clinical system, or working out where AI fits in a regulated setting, I’d like to hear from you.
There’s no form and no middleman: your email comes straight to me.
My CV: read it online ↗ or download the PDF ↓