Cornell Engineering's Scientific Artificial Intelligence (SciAI) Center has partnered with Pasteur Labs, an alumnus-founded startup, to break new ground in human-AI collaboration for scientific discovery and industrial applications.
The SciAI Center launched in 2023 with the aim of developing AI that can uncover fundamental scientific principles from large data sets using mathematics as a universal language. The partnership with Pasteur Labs, which specializes in AI simulation, will further that goal by establishing new research projects that reimagine the role of simulation in science and industry, said Christopher J. Earls, director of the SciAI Center and the J. Preston Levis Professor of Engineering.
"Whether it be for medical interventions or flying aircraft, decision makers require us to have confidence in the properties, precision and reliability of an AI system given the same repeated circumstances," Earls said. "What's great about Pasteur Labs is they have people who are savvy and understand the technologies and real-world use cases at a depth that we as academics might not."
Pasteur Labs was founded in 2021 by Alexander Lavin '12, an entrepreneur and expert in AI-for-science who said his company has a mutual interest in bridging what he calls the "valley of death" - the space between early technologies built bottom-up from theory and the need for those technologies to deploy in complex, real-world settings.
Computer simulations that attempt to model those settings have become vital testbeds for scientists and engineers. Coupled with advancements in high-performance computing and AI, these simulations can offer unprecedented insights in to fields such as climate science, materials and medicine, and, according to Lavin, new perspectives in high-dimensional data spaces and 'self-driving' laboratories that conduct fully-automated experiments.
"We can provide streamlined connections from research to practice and back by exposing the SciAI methods and academic projects to the noisy, messy environments - data set challenges, dynamic problem definitions, multidisciplinary constraints - that are pervasive in the real world," Lavin said. "In return, we're going to get feedback on our technology stack from some of the brilliant people at the SciAI Center, and they're going to influence the features that we put into production tomorrow."
One of Pasteur Labs' main business segments is advanced and additive manufacturing for aerospace and energy applications, where integrating data and domain knowledge in expensive AI simulation remains an industry challenge. The SciAI Center has three application focus areas - materials, turbulence and autonomy - which lend themselves to support of a multitude of manufacturing contexts, from semiconductors to molecular systems, and beyond.
"When it comes to these industries and computational technologies in concert with humans, we want to rethink what different types of scientific processes and instruments are now achievable," Lavin said, "and suddenly this very constrained space of human knowledge gathering expands several times, because we can query the natural world in unintuitive and rapidly more efficient ways."
The partnership brings together 50 faculty members, postdocs and doctoral students affiliated with the SciAI Center and 25 researchers at Pasteur Labs who specialize in multi-physics modeling, cooperative AI and machine learning. Among the first collaborators is doctoral student Harshwardhan Praveen, who is spending the summer conducting research at Pasteur's Brooklyn headquarters. Nikolaos Bouklas, SciAI Center co-investigator and associate professor of mechanical and aerospace engineering, will spend the next academic year on sabbatical leave at the company. Both Cornell researchers are contributing research to the company's launch of physics-AI applications in computer-aided engineering sectors.
Lavin, who called partnering with his alma mater an "unintentional but happy coincidence," reflected on his path from being an aerospace student at Cornell to an entrepreneur working with the university - a journey that included being named to the Forbes Science 30 Under 30 list and becoming an AI advisor for NASA.
"It's always surprising how small the world can be, especially in the scientific machine learning community," Lavin said, "But it's also a testament to the things Cornell does to encourage careers guided by the pursuit of knowledge."
