AI Product Leader
Doctor turned product leader. I build AI tools that help researchers discover what they need to advance human knowledge.
I started out as an NHS doctor. Those intense shifts showed me problems that were systemic, not just local, and I kept thinking about how technology could address them at a scale that clinical work alone couldn't. One side project led to another, and I ended up trading my stethoscope for product strategy.
I started out as an NHS doctor. Those intense shifts showed me problems that were systemic, not just local, and I kept thinking about how technology could address them at a scale that clinical work alone couldn't. One side project led to another, and I ended up trading my stethoscope for product strategy. Now I lead LeapSpace at Elsevier, the research-grade AI workspace built on Scopus data and publisher full-text. The work has changed, but the instinct is the same: help people access the knowledge they need to do their most important work.
The way I think about product work still comes from medicine. On the wards, you figure out what's actually wrong before you act, you watch what's happening rather than what you assume is happening, and you don't lose sight of who you're there for. That same discipline runs through everything I build, from the original Scopus AI prototype to LeapSpace, and open-source contributions like RAG-Fusion that others in the AI community have picked up and built on.
I think the interesting thing about going from medicine to AI for research is that the core problem is the same: people doing difficult, important work need the right knowledge at the right time. The waiting room just got a lot bigger.
Quick links to discover more about my work
My journey from NHS doctor to product manager
Where I've worked and what I've built
AI tools and side projects I've created
Writing about product management and AI
Talks, podcasts, and appearances
Highlights from a decade of reading
Let's connect and collaborate
Handpicked pieces I'm particularly proud of or that sparked the most interesting conversations
and why I've been making Bob Dylan songs about Sonic the Hedgehog
and why I've been making Bob Dylan songs about Sonic the Hedgehog
and how to use tarot cards for sprint planning.
and how to use tarot cards for sprint planning.
Building Faster, More Reliable Agents with Sequential Retrieval-Augmented Generation
Building Faster, More Reliable Agents with Sequential Retrieval-Augmented Generation
Favourite highlights from a decade of reading, captured with Readwise
The test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to function.
F. Scott Fitzgerald · The Crack-Up
Product work is full of contradictions. This one reminds me it's fine to sit with them.
The test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to function.
Product work is full of contradictions. This one reminds me it's fine to sit with them.
F. Scott Fitzgerald
The Crack-Up
We are what we repeatedly do. Excellence, then, is not an act, but a habit.
Will Durant · The Story of Philosophy
I come back to this whenever I'm tempted to chase a shortcut instead of building the muscle.
We are what we repeatedly do. Excellence, then, is not an act, but a habit.
I come back to this whenever I'm tempted to chase a shortcut instead of building the muscle.
Will Durant
The Story of Philosophy
The chief enemy of creativity is good sense.
Pablo Picasso · Conversations with Picasso
A useful antidote to the part of my brain that wants every idea to be sensible before it's even born.
The chief enemy of creativity is good sense.
A useful antidote to the part of my brain that wants every idea to be sensible before it's even born.
Pablo Picasso
Conversations with Picasso
Things I keep turning over. No answers yet, and that's the point.
What would academic research look like if we designed discovery tools for curiosity rather than productivity?
How do you build trust in AI systems when the people using them understand the domain better than the system does?
Is there a version of 'evidence-based' thinking that doesn't accidentally filter out the questions worth asking?
What's lost when we optimise for finding answers instead of sitting with better questions?
You found it! This is what happens when you give a doctor access to 3D scanning equipment. No patients were harmed in the making of this digital twin.
Tap Interrupt to explore