Why I’m joining Astera to lead Radial

The systems that enable how we fund, do, and build upon science have long needed an update.

Academia is often caught between training students and chasing publishable novelty. That dual mandate makes it hard to prioritize long-range, systems-level work with no obvious paper at the end of it. In addition, the winner-takes-all incentive structure tends to discourage true teamwork.

Industry can do extraordinary translational work and attract top talent, but is structurally disincentivized from basic research, tool building, or sharing data that could benefit the broader field.

Venture capital is a powerful accelerator, but its logic doesn’t stretch to ideas that are too long-term, too coordination-heavy, or don’t fit a traditional high returns-generating business model. As an investor in biotech platform companies I continually saw amazing technologies get built, but their impact stunted by the necessity to focus on the first commercially-viable output.

The result is a missing middle: a real gap that existing funding structures aren’t designed to touch. Ideas that are too unglamorous to satisfy a grant or journal, too ambitious for a 10-year VC return, and too cross-functional for a siloed lab or individual company.

I’ve seen all of this first-hand. Going through a PhD and postdoc gave me an early sense of how much work it takes to go after true unknowns, and all the bottlenecks our academic system inadvertently generates. Biotech industry roles spanning research to business development allowed me a first-hand view of what it actually takes to move science from a discovery to something useful in the world. In venture, I had the unique privilege to build alongside some of the most creative scientists and founders tackling some of the most ambitious problems in life sciences.

Across every experience, I’d encounter science that absolutely should exist and yet realize there was nowhere for it to go. Great ideas with clear impact, but the timelines were too long, the coordination too complex, or the commercial logic too uncertain to ever make a defensible funding pitch.

But now, I have a chance to actually do something about the problems that I’ve observed first hand. I am joining the Astera Institute to lead Radial because I believe we can finally build a home for those ideas, and reimagine life science research systems in the process.

I’m drawn to Astera because of our express mission to not only take big bets, but to build useful things. Philanthropy has the potential to play an outsized role in accelerating science when structured the right way, and Astera is taking an approach that offers enormous flexibility and supports work that operates with the focus and speed of a startup. This structure allows us to experiment on science’s systems-level challenges by actually doing ambitious research on root-node scientific questions, rather than narrowing to a specific disease or research area.

AI makes addressing the gap more urgent and possible

Now is also an ideal time to do this. Running alongside the institutional challenge is a complementary one: to realize the promise of AI for science, we have to change how we feed the machine. AI can operate at the scale and speed needed to actually learn from the full volume of scientific experiments and observations, but only if we fundamentally redesign how that data is generated, shared, and used.

The scientific system as it currently exists wasn’t built for that. Most data gets generated without reuse in mind, experiments never get shared in a form that the next person (or machine!) can actually build on, and funding cycles reward the wrong timelines. These aren’t new problems, but AI makes the cost of ignoring them much higher.

The flexibility to do something about it

Radial is well-suited to tackle these systems level challenges for a few reasons. First, we will fit the funding to the problem, not the other way around. We have the operational muscle and flexibility to do what the market can’t, whether it’s building in-house, orchestrating public-private partnerships, making for-profit investments, or giving a researcher unrestricted space to let an idea take shape.

Second, we are taking a “full stack” approach. The biggest bottlenecks in science today are the result of a lack of coordination, not a lack of talent. Radial is designed to fund and coordinate across the full stack, from basic discovery to the unglamorous but important work of building infrastructure and tooling.

And third, we’re embracing a culture of experimentation. If we aren’t running experiments with a high risk of being wrong, we aren’t reaching far enough. Our goal is to learn in public and generate science that’s genuinely useful and reusable.

What this looks like in practice

We’re already building, taking multiple bets in different places where the system most needs to change. Radial’s first program, The Diffuse Project, aims to unlock our understanding of protein dynamics at a scale never before possible. To start, we are redesigning the entire X-ray crystallography pipeline from sample prep to data collection to modeling, capturing information about protein motion that standard workflows routinely discard. The project is coordinated across full-time Radial team members, multiple universities, and national labs, with everything released openly in real time (check out this awesome video explainer!).

Alongside Diffuse, we’re beginning to build The Stacks, a new experimental publishing platform built from first principles for machine readability and genuine reuse rather than retrofitted from journal conventions. We believe that changing the way that scientific information is shared and discovered is essential for realizing the full potential of AI’s impact in science.

We’re also expanding our support for OpenADMET, a public-private partnership aimed at producing open, high quality datasets on small molecule properties. Our additional funding will enable them to build an AI model competition platform so the field can actually reuse and build on the work. This reflects our interest in collaborating with start-ups building unique platform technologies to create public good datasets, and in funding the unglamourous infrastructure needed to generate science that’s genuinely useful and reusable.

This is all the kind of work I would have loved to fund in a past life but couldn’t, and I’m really excited to support it here.

The unknown as a feature, not a bug

We’re not pretending we know exactly how to solve the challenges we’re confronting today. We don’t even know all the right questions to ask yet. But honestly, the “open-endedness” of our mission is what I find most exciting.

I feel incredibly lucky to have the space and the resources to take some bold bets, do the experiment, and actually try to build something that doesn’t exist yet. If this idea excites you too, reach out! LFG!!