Q&A with chemist Ilya Shkrob on autonomous discovery at Argonne

Ilya Shkrob has spent decades as a chemist at the U.S. Department of Energy’s (DOE) Argonne National Laboratory, working at the interface of chemistry, physics and advanced instrumentation. Today, he is helping to build a new kind of laboratory — one where experiments run themselves. His team has launched an autonomous discovery lab, combining robotics, artificial intelligence (AI) and advanced data analysis to accelerate scientific research. In this conversation, Shkrob explains how autonomy can transform scientific discovery, what makes Argonne uniquely equipped to lead this work and why this shift could help solve some of society’s most urgent challenges.

“Autonomous doesn’t mean the system has intent or judgment the way a human does. It means the system can operate independently: adapting to data, learning from outcomes and guiding the next steps. That independence opens up exciting new possibilities for accelerating research.” — Ilya Shkrob, physical chemist

Q: What does ​“autonomous discovery” mean in the context of scientific research?
A: It refers to the use of robotics and machine learning (ML), which is an application of AI, to carry out key parts of the research process — designing and running experiments, collecting data, analyzing results and even deciding what to try next. The goal is to reduce the need for constant human intervention and make the entire process faster, more precise and more scalable.

Autonomous doesn’t mean the system has intent or judgment the way a human does. It means the system can operate independently: adapting to data, learning from outcomes and guiding the next steps. That independence opens up exciting new possibilities for accelerating research.

By flagging unusual results and surfacing outliers, for example, autonomous systems can point researchers toward phenomena we might otherwise overlook. That’s a valuable complement to human intuition and critical thinking. The goal isn’t to replace human creativity, it’s to enhance our ability to spot what matters.

Q: What types of research problems are best suited for this approach?
A: Autonomous systems help us explore more ideas, faster. They’re especially valuable when we need to run large numbers of experiments or gather data continuously over long periods. By handling routine, repetitive or highly precise tasks, these systems free researchers to focus on strategic questions and creative insight.

One recent example from our Robotic Autonomous Platforms for Innovative Discovery lab (RAPID-200) for materials and chemistry involved redox flow batteries — systems that store energy in tanks of fluid and discharge that energy on demand. To be practical, the molecules in these fluids must remain stable for thousands of hours.

We ran thousands of experiments using our autonomous system. That’s a scale no human team could manage alone in a reasonable amount of time. We discovered some very promising candidates, but just as important, we found that past a certain point, no further improvements were possible. That told us there’s a fundamental limit to what these molecules can do under current conditions. Knowing where the boundaries are is just as important as finding new materials. It helps us ask better questions and focus our efforts on more promising directions.

Q: What makes your autonomous discovery lab different from others?
A: One key distinction is that RAPID-200 is designed to study how systems evolve over time. Many autonomous labs focus on rapid screening — experiments that give immediate results and don’t change much over minutes or hours. But in fields like energy storage and quantum technologies, degradation and time-dependent behavior are everything.

RAPID-200 is built to monitor reactions, materials and devices continuously — over hours, days or weeks. That allows us not only to capture long-term performance, but also to adapt the experiment as it progresses. That time-resolved approach opens a much wider range of scientific questions and makes our lab special.

Q: Why is Argonne an ideal place to pursue autonomous discovery?
A: Argonne brings together all the right elements for autonomous discovery to thrive: important national challenges, multidisciplinary expertise and a culture of collaboration. We’re working on problems that matter — energy storage, critical materials, chemical manufacturing and quantum systems. These are areas where traditional experimentation is too slow and where autonomy can dramatically accelerate progress.

Argonne is also home to world-class DOE Office of Science user facilities such as the Advanced Photon Source, the Center for Nanoscale Materials and the Argonne Leadership Computing Facility. These resources allow us to combine high-throughput experimentation with advanced imaging, modeling and simulation at a scale that few other institutions can match.

Equally important is the human side. We have an exceptional community of chemists, physicists, engineers, computer scientists and robotics specialists — all working together. Autonomous discovery doesn’t belong to a single discipline. It requires the convergence of hardware, software and scientific insight. Argonne has that mix.

Finally, our leadership understands the importance of transforming how science is done. They’ve supported the development of RAPID-200 and other autonomous discovery labs not as demonstrations, but as serious platforms for breakthrough science. That vision — and the willingness to invest in it — is what sets this place apart.

Q: What are some of the practical challenges in building and operating an autonomous lab?
A: One of the biggest challenges is logistics. If you want to test 10,000 materials, you have to synthesize 10,000 samples, track them reliably, process them efficiently and ensure that the data you generate is accurate and useful. That’s not just a matter of designing good experiments — it requires building a robust, end-to-end system to support every step. You need automation across the entire workflow: from sample preparation and experimental setup to measurement, data management and decision-making. Scaling that up takes a combination of engineering discipline and well-integrated software.

Another challenge is robustness and reliability. A truly autonomous system must be able to run continuously for days at a time without human intervention. Once we reach that level of stability, we can begin to tackle more complex and open-ended scientific problems with confidence.

Q: What will be the broader impact of autonomous discovery on the scientific community?
A: I believe it will fundamentally change how we approach research. Autonomous systems allow us to test more ideas, faster and with greater consistency. That will mean we can ask questions that were previously out of reach because they required too much time labor or data.

In the long run, I believe these tools will make high-level science more accessible to members of the scientific community. Labs that once needed large teams to tackle big problems may be able to accomplish more with smarter infrastructure and fewer people deployed to any one problem, freeing scientists to confront more challenges at once. That’s a win for science and society.

Q: Why is this kind of work important now?
A: The world has many pressing problems and needs solutions faster than ever. Autonomous research is one way to help meet those needs. It gives us a platform for generating knowledge more quickly, testing technologies more efficiently and refining scientific understanding on a faster timeline.

Our goal in the RAPID-200 lab is to support national science and security priorities. We’re not just trying to automate what scientists already do — we’re trying to build a new model for how science can be done at a national scale.