PrimaLabs joins NVIDIA and national labs to advance the DOE's Genesis Mission

The Genesis Mission

PrimaLabs was built by people who spent 15 years inside the Department of Energy's national laboratories, extracting maximum performance from the world's largest machines. Today, that work comes full circle: PrimaLabs, led by co-founders Prasanna Balaprakash and Chaitanya Hiremath, has been selected as an industry partner on two Savannah River National Laboratory-led projects under the U.S. Department of Energy's Genesis Mission.

The Genesis Mission, launched by executive order in November 2025, brings the DOE's 17 national laboratories together with industry and academia around a single goal: to double the productivity and impact of American science and engineering within a decade. The DOE selected 278 projects through a rigorous merit-review process, following the largest response to a competitive funding opportunity in the department's history.

For us, joining the Genesis Mission is personal. Our co-founder and CEO, Prasanna Balaprakash, served as Director of AI Programs at Oak Ridge National Laboratory, where he led research that helped lay the groundwork for the Genesis Mission itself, including the model consortium that now anchors its AI portfolio. Returning to the mission as an industry partner, this time with PrimaLabs, is exactly the kind of consequential problem we started the company to work on.

The Two Projects

Both projects tackle a challenge most people never think about, but every American has a stake in: safely processing the nation's legacy nuclear waste.

VITA-SCALE

VITA-SCALE will use physics-informed AI to translate laboratory and pilot-scale vitrification data into predictions for full-scale nuclear waste processing. Vitrification, the process of immobilizing radioactive waste in glass, is one of the most complex industrial operations in the world. Small-scale experiments are abundant, but confidence at full scale is not. The goal of VITA-SCALE is to close that gap.

SCOPE

SCOPE will combine accelerated chemistry modeling with AI to predict the behavior of radioactive salt waste before it disrupts treatment operations. When chemistry surprises a waste treatment facility, the result is downtime measured in months and cost measured in tens of millions. SCOPE is built to see those surprises coming.

Together, the projects bring PrimaLabs, DOE National Laboratories including Savannah River National Laboratory, Oak Ridge National Laboratory, and Pacific Northwest National Laboratory, and NVIDIA into a single effort to improve safety, lower costs, and accelerate scientific and operational decision-making across one of the nation's most demanding environmental cleanup programs.

Where PrimaLabs Contributes

The problem underneath both projects is one we know well: getting the most out of fixed GPU hardware under hard constraints, where every hour of compute has to count. That is the same discipline behind PrimaLabs as an AI inference platform and AI serving platform — GPU-optimized inference engineered so teams can reduce AI inference cost without sacrificing performance.

That is the discipline PrimaLabs was founded on. Our team trained the largest AI models ever run on Frontier, the world's first exascale supercomputer, including ORBIT, a 113-billion-parameter foundation model that sustained 1.6 exaFLOPS across 49,152 GPUs and earned back-to-back ACM Gordon Bell Prize Finalist recognition in 2024 and 2025. The same self-adaptive approach that powers our commercial AI inference engine — observing the live workload, learning what works, and tuning across the serving stack continuously — is what these scientific workloads demand. Physics-informed models and accelerated chemistry simulations are not static benchmarks. They are constantly evolving workloads, and the stack that serves them has to adapt to each one. That is LLM inference optimization applied where the stakes are national.

Static stacks are optimized once. Ours learns every workload. That is true whether the workload is a coding agent serving millions of requests or a physics model predicting how nuclear waste processing behaves at full scale.

Why This Matters

Scientific AI is moving from demonstration to deployment. The Genesis Mission is the clearest signal yet that the federal government intends to run AI at national-lab scale, on real problems, with real operational consequences. Working alongside Savannah River National Laboratory and NVIDIA puts PrimaLabs at the center of that shift, applying the performance engineering behind our platform to problems where the stakes are national.

We could not have asked for a more meaningful place to bring the two halves of our story together: the national-lab heritage that shaped how we build, and the adaptive inference stack we are shipping to customers today.

The official Genesis Mission announcement is available at genesis.energy.gov.

If applying serious performance engineering to problems that matter sounds like your kind of work, we are hiring. Or just say hello: [email protected].