US science is taking hits: is Julia affected?

Will US science survive Trump 2.0?

How are Julians affected? Has the funding dried up for commercial Julia activities? NIH, for example, has severely cut budgets left and right…

7 Likes

Also of interest: pack your bags and move to Europe… Several countries are ready for you.
A brain drain would impoverish the United States and diminish world science

1 Like

I described a lot of the funding for Julia infrastructure in this talk:

It’s specifically titled “Building a product which improves open source sustainability” because it details this strategy of funding open source through commercial ventures built on the open source tools.

I’ll stress that it’s not all of it of course, but it covers the Julia Lab, JuliaHub, and related entities and walks through the stories of their evolving funding mechanisms. One of the main things that is stressed is that, continually as we have been growing, Julia has had more of its core contributions being funded through commercial entities, in particular PumasAI and the JuliaHub work around JuliaSim (soon to be renamed and launched… details to come soon :wink:). This was done already because the uncertainty around federal funding has always existed, it’s now just testing everyone.

Now this talk was given 3 months ago, so things have clearly changed and become more uncertain, but it gives us a window into the potential opportunities and the pathways to go. Increasingly, we cannot rely on large NSF or NIH grants to fund a large MIT Julia Lab that builds the next generation of open source maintainers. We will still be going for whatever good funding mechanisms exist (in particular, the NSF CSSI grants have always been a good fit), but with these agencies cutting their budget by 50%, we should effectively assume that the Julia Lab is going to shrink by 50%.

That said, there are now more opportunities in terms of SBIR/STTR mechanisms, so increasingly one of the mindsets that I want to be fostering is helping late-stage students and postdocs enter an entrepreneurial mindset and use funding resources that require a path to commercialization. One such project is Neuroblox, which for a few years was an MIT/Harvard/Stonybrook collaboration but is now increasingly looking to perform this transition and is looking for grant pathways with DARPA/ARPA-H/SBIR/STTR and a path to commercialization of the use in the pharmaceutical and device industry. Some of the efforts around cardiac modeling is taking a similar path.

In fact, we are rather fortunate that the core business of JuliaHub has been to build a platform that is the perfect environment for bring Julia code into regulated industries: it is a SOC2, HIPPA, and CFR Part 11 compliant platform that makes so that applications written with a “good enough” software development lifecycle (SDLC) satisfies the regulatory constraints for being used in pharmaceutical contexts, and increasingly we’re covering aspects for things like automotive and aerospace. So this X+JuliaHub partnership is then a strong path quickly getting some traction in regulated industries with scientific needs that gives us a new playbook that isn’t reliant on continued government support.

Now I have to say, I obviously love open source. I can guarantee that I have written more open source software than the vast majority of people who will ever read this post. But I think we can use the learnings from Pumas, where at first I had wavered in the idea of it going commercial, but now it very clearly is a strong and sustainable funding entity which has hired many of the top contributors to the open source Julia ecosystem and has given them a day job which extends their work. It has been a success not just in building a company but also as a model for open source sustainability that has a positive mindset: people hired there understand they will now be building a product, but it’s built on open source tools they contribute to and thus a bit of both continues and we all understand where the money comes from in order to keep the bricks on the bottom standing.

Now you are seeing that for example some of the motives that are driving the major advancements of Julia v1.12 as the needs in order for JuliaSim to be put together. A lot of the work on making small binaries for Julia to ship on embedded platforms has stemmed from commercial needs where the hands that have made these contributions are parts of the teams related to shipping a product for industrial modeling and simulation with a feature for code generation to real-world devices. There will always be questions of “what should be open sourced and what shouldn’t in order to have a viable business model?”, but hopefully this showcases that it does not need to be binary thinking.

Call to Action

So let me end with a call to action. If you are needing to seek a new path to funding and would consider transitioning from academia to industry as a new way to sustain your Julia contributions, feel free to get in touch. There are many models that could work. We could partner on SBIR/STTR grants, build partnerships through JuliaHub, or just talk about commercialization models since we now have quite some experience building up commercial software products built on open source. We are talking with people doing not just pharmaceutical related things but also domains like power systems and autonomous flight. I’m happy to get creative and find new pathways.

Final Thoughts

I want to end by showing how to look at the world in a positive mindset. But there’s so much opportunity if you keep an open mindset. Take this FDA memo for example:

Some people have said it’s reckless: we need animal testing but we kind of need it right now! Instead it says we should:

  • Advanced Computer Simulations: The roadmap encourages developers to leverage computer modeling and artificial intelligence to predict a drug’s behavior. For example, software models could simulate how a monoclonal antibody distributes through the human body and reliably predict side effects based on this distribution as well as the drug’s molecular composition. We believe this will drastically reduce the need for animal trials.

It’s asking for everyone to use a software that doesn’t always exist! It’s pushing the requirements to have everyone use a software that doesn’t currently exist… let’s make it?

53 Likes

Thanks for the detailed response @ChrisRackauckas , this is very informative and helpful in many different ways. I would encourage folks to look through this resource as it seems quite valuable for junior faculty even without considering the current federal situation.

@PetrKryslUCSD , could you consider renaming this post? I actually find it unnecessarily inflammatory of a title. The USA might be sliding down into an authoritarian hellhole, but there are many people still working hard to protect its institutions and talent, trying to weather this catastrophe and undo its effects. Science work in the US has certainly become more difficult, but it is continuing, and very much not “down the drain” (yet).

4 Likes

Stefan, thank you for your comment. Alas, stating the true state of affairs may be inflammatory – people actually care what happens to their science, and get understandably upset when it starts circling the drain. I know first hand of several colleagues in federal research agencies of good repute, who are currently desperately searching for jobs as they expect to get fired any day now. My own university will start letting researchers go, as it may experience federal research fund cuts of the order of $500M. And so on.

Note for instance
Science under attack! - Offtopic - Julia Programming Language

9 Likes

I admire your upbeat enthusiasm, Chris. Alas, even things that look like opportunities today may evaporate tomorrow. The NIH got hammered so thoroughly that the aftereffects may spoil things for years to come. For those interested, watch the bit about NIH:
RFK Jr. & HHS: Last Week Tonight with John Oliver (HBO)

1 Like

That is very interesting, thanks for letting us know about this.

I wonder how much attention has been paid to the need for validation.
For that one must have experimental data from physical (animal) experiments.
Also, uncertainty quantification will be absolutely key. Any thoughts?

2 Likes

In the context of the wider thread, it may appear that this is among the reckless decisions of the current administration (I’m putting that as nicely as possible), but this was actually passed by the previous administration. The exact wording is important here: it’s not phasing out all animal testing, just the mandate by the original 1997 law. While it’s indeed reckless to dismiss animal testing because “90% of animal-tested drugs fail in human trials anyway” (think of all the poisons animal testing stopped from reaching human trials, aka real people in a tough situation), there is a valid point that we don’t necessarily have useful animal models for everything, and we could potentially get the same information with other emerging computational and biological technologies in some cases.

It’s worth highlighting that the biological side will likely be more important than the computational side in the near future. Computational biology is making great strides in research e.g. AI-driven protein structure prediction, but the computer’s knowledge of human biology is about as limited as ours. It won’t model a biological pathway we don’t tell it to, and we can’t tell it if we didn’t discover enough of it in the first place. For another example, we can do a decent job at modelling particular protein-ligand interactions in favorable conditions (which doesn’t automatically follow from accurate structure prediction), but we can’t feasibly compute across every known and unknown protein in the human body, let alone infer the physiological consequences in real time. Contrary to what some animal rights activists are celebrating, we realistically will still see a lot of animal testing because injecting a drug into a somewhat-off black box can be faster, cheaper, and more reliable than wrong computational predictions wasting more human trials (remember, 90% accurate is 10% wrong).

More relevant to the main topic, America’s ability to advance any of these technologies has been hampered by the attack on the public funding and political independence of academia. Industry tolerates basic research less due to the increased pressures of shorter term financial gains, and that will hurt some fields far more than others. I’d say that resistance is just as important as adaptation in optimism: oppose rising anti-intellectualism and the associated politics because besides publicly funded basic research being foundational, anti-intellectuals don’t actually care whether you and your colleagues are doing research for a university or a company.

9 Likes

The other species that have lived on this planet for millions of years, most of them occupying the house long before us and allowing us to exist, have no right? We conveniently ignore the fact that we are all primates, evolutionary cousins of the other mammals, and that like them we are all mortal animals. I just hope that a more advanced species that visits us one day has higher moral principles than ours (yours), because if not, I don’t want to imagine what a nightmare it would be.

3 Likes

I didn’t make the argument you’re suggesting I made. It’s just a fact that the new legislation in effect does not ban or discourage animal testing, and it probably won’t make as much of a difference as animal rights activists predict. The legislation certainly didn’t give animals more rights.

4 Likes

I guess Julia being affected is a small problem overall…

https://www.nature.com/articles/d41586-025-01396-2

That and what’s happening to Harvard will probably require a more incisive position of all the accademic institutions.

7 Likes