The Hidden AI Bottleneck: Why PhD Talent Matters

Bill Kerr, founder and CEO of Athyna, believes the next phase of AI progress will be shaped by both compute and human expertise. While advances in hardware remain essential, scaling AI systems into production increasingly depends on access to PhD-level researchers, especially in post-training. This is the stage where models are evaluated, refined, and aligned for real-world use, and where scientific judgment matters as much as raw processing power.

That's why Athyna is launching Athyna Intelligence. Building on years of experience hiring, vetting, and supporting elite global talent, this new platform connects AI Labs and Fortune 500s with deep scientific talent across Latin America, bringing PhD-level expertise directly into post-training, evaluation, and alignment workflows—exactly where AI development needs it most.

We spoke with Kerr about the growing shortage of advanced AI talent in the U.S., why post-training has emerged as a major bottleneck, and how global expertise, particularly from Latin America, is beginning to reshape how frontier AI teams are built.

AI has never attracted more attention or capital. Why do you think talent has become one of the most important limiting factors now?

If you look at how AI systems are actually being deployed today, you start to see a pattern. Pre-training gets you impressive general capabilities, but it doesn’t get you reliability, reasoning, or trust. That work happens later, in post-training.

Post-training, models are evaluated, stress-tested, aligned, and adapted for real-world use cases. And that phase is fundamentally human-driven. It depends on people who understand math, logic, science, language, and law—the underlying structures of the problems AI is being asked to solve.

The issue is that the U.S. simply doesn’t produce enough PhD- and Masters-level researchers to support that work at the scale the industry now demands. Demand has exploded, but the talent pipeline hasn’t.

AI skill demand is shifting toward deep technical and scientific expertise.

That surge is visible in hiring data as well. Recent LinkedIn analysis shows that AI Engineer, AI Consultant, and AI/ML Researcher roles now rank among the five fastest-growing positions in the United States. These are roles that were marginal or nonexistent just a few years ago, underscoring how quickly organizations are shifting from AI experimentation toward the need for deeper technical and research expertise.

You mentioned post-training. For readers less familiar with the term, why is it so important?

Post-training is the difference between a model that looks good in a demo and one that works safely in production.

This is where models learn to reason through complex problems, handle edge cases, follow instructions reliably, and avoid failure modes. It includes evaluation, alignment, reinforcement learning from human feedback, and domain-specific refinement.

None of that can be done effectively with generic annotation. It requires expert judgment. People who can look at a model’s output and understand why it’s wrong, not just that it is. That’s why advanced academic training matters so much here.

If the U.S. talent pipeline can’t keep up, where should companies be looking?

They should be looking globally, and, more specifically, south. Latin America is one of the most untapped talent regions in the world. The region produces tens of thousands of PhDs and Master's graduates every year across mathematics, physics, computer science, biology, and law—exactly the disciplines that post-training depends on.

These researchers are publishing internationally, collaborating with top universities, and working in US-aligned time zones. The difference is that the market hasn’t built the infrastructure to fully integrate them into frontier AI work at scale.

Is that gap what led to the launch of Athyna Intelligence?

Exactly. Athyna Intelligence is a natural extension of what we’ve been building for years. Athyna has long focused on sourcing, vetting, and operating global talent, particularly across Latin America. We’ve built AI-powered systems to quickly match thousands of talents across hundreds of companies.

With Athyna Intelligence, we’re applying that expertise specifically to AI post-training. The goal is to make PhD-level expertise accessible on demand, with fast turnaround, so AI teams can move at the pace their work requires.

How does Athyna Intelligence actually work in practice?

We start with people. Everything else follows from that. Our first step is connecting companies with vetted PhD and Masters-level researchers from Latin America who are available on demand and ready to contribute quickly. These are not generalists; they’re domain experts with the academic depth this phase of AI development requires.

Around that talent, we will build the operational layer: workflows, review processes, and execution models that turn expert judgment into consistent, repeatable output. We shape those workflows closely with design partners, using real production needs to guide how work is structured and iterated. The focus is always on speed, rigor, and real-world usefulness.

What kinds of challenges are you seeing that signal strong demand for this approach?

We’re seeing the same issues keep coming up. As models improve, evaluation shifts from volume to precision. Teams can generate massive amounts of evaluation data, but they struggle to find enough people who can correctly interpret failures—especially in domains where a wrong answer can appear superficially correct.

Models struggle with reasoning-intensive domains such as math, physics, and causality. Safety and alignment teams are overwhelmed because red-teaming requires intent and expertise, not trial-and-error.

And evaluation pipelines often break down because organizations don’t have enough qualified people to run them properly. The industry needs better data, and more importantly, it needs the right people generating and evaluating that data. That’s the gap Athyna Intelligence is designed to close.

What does this mean for researchers in Latin America?

For years, highly trained researchers in Latin America have faced a difficult tradeoff: leave the region to work on cutting-edge AI systems, or stay and apply their skills in roles that rarely match their level of training. Athyna Intelligence introduces a third path.

Researchers can now contribute directly to frontier AI systems from where they live, earn globally competitive compensation, and apply their expertise to some of the most technically demanding problems in the field. This access benefits companies building advanced AI, but it also reshapes who gets to participate in that work.

The result is not only better AI systems grounded in deeper scientific judgment, but also a broader, more diverse group of researchers helping shape how those systems behave in the real world. For many PhD-level scientists in the region, this represents a rare opportunity to work on technologies that will define the next decade, without relocating or stepping away from research-grade problems.

Thank you, Bill, for the conversation.

To learn more about Athyna and how Athyna Intelligence supports organizations working on advanced AI systems, you can explore more information here: https://www.athyna.com/intelligence

—

*Sponsored by Athyna. We have equity in the company.

Disclaimer: The Bay Area Times is a news publisher. All statements and expressions herein are the sole opinions of the authors or paid advertisers. The information, tools, and material presented are provided for informational purposes only, are not financial advice, and are not to be used or considered as an offer to buy or sell securities; and the publisher does not guarantee their accuracy or reliability. You should do your own research and consult an independent financial adviser before making any investments. Neither the publisher nor any of its affiliates accepts any liability whatsoever for any direct or consequential loss howsoever arising, directly or indirectly, from any use of the information contained herein. Assets mentioned may be owned by members of the Bay Area Times team.

Please read our Terms of Service and our Privacy Policy before using Our Service.