Hello, I’m Arfon

A former astronomer, I co-founded the Zooniverse and created the Journal of Open Source Software, then founded and led the Data Science Mission Office at STScI, home to the Hubble, James Webb, and Roman Space Telescopes. I was later a Director of Product for Copilot at GitHub and most recently a Senior Fellow at Schmidt Sciences. I now work on making AI work better for science at Fulcrum.

In many domains, faster science is a good thing. Unverified science is not. My work focuses on three themes:

🚀 Building for successful AI adoption in science: the people, skills, and tooling researchers need so that AI delivers real knowledge, not just more output. That includes capturing the domain judgment and expertise that models can’t generate for themselves.

🔬 Connecting AI to the physical world: much of science happens at the bench, not in code. I’m interested in lab automation, instrument interfaces, and simulators that shorten the loop between an AI-generated idea and a real-world result.

⚡ Trustworthy, verifiable science: AI has made producing research dramatically cheaper, but checking it hasn’t kept pace, and the old trust signals no longer mean what they used to. I’m interested in making verification an engineered, open part of how science works, from benchmarks and datasets to provenance that travels with results.

Before Fulcrum, I was a Senior Fellow at Schmidt Sciences on the Science Systems team, and before that a Director of Product at GitHub, where I led efforts to integrate AI into the core GitHub product experience.

I’m an experienced leader and strategist with deep product, data science, and engineering experience, and a particular focus on open source. I believe the hardest problems require close partnership between product, design, (data) scientists, and engineers, and my superpower is speaking the language of all of them. I’ve worked at organizations of very different sizes, scales, and sectors, from volunteer-run open source projects to GitHub and NASA flagship missions.

A lapsed academic with a passion for new models of scientific collaboration, I’ve used big telescopes to study dust in space, built sequencing platforms in Cambridge, and helped build the Zooniverse into a platform engaging millions in online citizen science.

I’ve been known to publish the odd paper and have also built a couple of academic journals, including the Journal of Open Source Software (JOSS), a developer-friendly, open-access journal for research software packages.

Contact

If you need to contact me then say hi on Bluesky, find me on GitHub, or send me an email.