Three IBMers Named to MIT Technology Review's 35 Innovators Under 35

They are inquisitive and persistent. They take the road not taken. They are inspired and inspiring. They question. They experiment.

Meet three members of IBM's Research community who are part of the elite class of the 2020 MIT Technology Review's 35 Innovators Under 35 honor roll—Jennifer Glick, Zlatko Minev, both of IBM Quantum, and Manuel Le Gallo-Bourdeau, a researcher in neuromorphic and in-memory computing.

Selected by their peers, the publication annually profiles young scientists for their commitment to innovation and using technology for good. The tenacious trio from IBM are changing the future of science and technology with vigor and persistence.

Jennifer Glick

Asked what she finds so surprising about the quantum computing industry, Jennifer Glick answers that it's the rapid, constant growth the industry has seen, even in its infancy of a technology that is poised to be a game changer. "In particular, the companies that have been leaders in their industries by partnering with us to do applications research," said Glick who spends her days working on IBM's QStart program, guiding clients on real-world quantum applications to help their businesses.

Glick first became fascinated with quantum computing when she was studying physics at Michigan State University where she earned her doctorate, and she took note of how different tech companies approached the nascent technology, especially IBM. "I wanted to be part of something from the ground up," she said. "And with something where I could apply my physics training in something that had would have a wide impact in industry and society."

And Glick has been front and center of that change with her job focusing on how members of the IBM Q Network (which include more than 100 Fortune 500s, academic institutions, research labs, and startups) – with interests ranging from chemistry, finance, aviation and AI – can develop commercial applications with IBM's quantum technology.

Quantum computers have only been accessible outside of the lab for a few years and Glick works with clients to make connections between abstract quantum algorithms and the challenges they face.

For example, the 2019 paper she co-authored with colleagues from IBM and the investment bank, Barclays, "Quantum Algorithms for Mixed Binary Optimization applied to Transaction Settlement," introducing a new approach for extending existing quantum methods for combinatorial optimization. The research demonstrated quantum optimization of settlement efficiency, a process crucial to finance market infrastructure due to the volume and value of transactions settled. Better quantum tailored algorithms could increase settlement efficiency, minimizing the time period between trade and settlement.

Glick has also shared her expertise at industry events, such as JPMorgan Chase's Innovation Week and TechFest, and The New Lab Quantum Summit 2019, where she made presentations on quantum computing and quantum machine learning algorithms. For her efforts to help businesses explore ways of mapping industry-relevant problems to algorithms that can be executed on quantum hardware, she has earned IBM's Outstanding Technical Achievement award. She has also taught quantum machine learning using Qiskit, IBM's open-source quantum computing software development framework, to industry experts at IBM Think 2019, as well as early adopter students and developers at IBM-sponsored workshops and boot camps.

Glick's advice for future researchers who want to be part of the quantum community? "I would tell them to start learning as much as you can right away. We have Qiskit which is a good way to start now, working on real hardware. I would also seek out research and internship opportunities in quantum, especially in industry. I think that's a really invaluable experience for young people in school."

Zlatko Minev

Zlatko Minev grew up fascinated with the breakthroughs of Nikola Tesla and the ability to solve problems deemed impossible by unearthing and seizing the power of new science. While taking summer college classes at Stanford University during his high-school days, he made his life choice to pursue physics — in his view, the most fundamental science that can uncover nature at its deepest level. Pursuing physics in college at UC Berkeley, through a series of unexpected events, a freshman Minev found himself working in one of the early quantum computing labs.

And the rest is history.

Since then, Minev has been making 'quantum leaps.' Literally.

Minev surprised the quantum community by settling a debate that's divided physicists for over a century — it was assumed that the energy level of atoms changed in abrupt, random, so-called 'quantum jumps.' Minev devised a landmark experiment and advanced quantum technology to show the contrary — in an experiment deemed impossible. His experiment not only predicts the occurrence of quantum jumps but seizes control of their imminent fate. Last year, Minev published these results from his Ph.D. dissertation at Yale in an article in Nature magazine, "To Catch and Reverse a Quantum Jump Mid-Flight."

Nature summarized the significance of the results: "[The] experiment overturns Bohr's view of quantum jumps, demonstrating that they possess a degree of predictability and when completed are continuous, coherent and even deterministic." The discovery has the potential to transform our fundamental understanding of quantum physics, and was selected as the top Math and Physical Sciences discovery of the year by Discover Magazine.

Technologically, Zlatko's innovations led to the most sensitive and time-resolved quantum measurements and feedback to-date, paving the way to greatly improved quantum sensing technology. Combined with his new results on predictability, these open the door to radically new technologies to correct the errors be deviling quantum computers.
Minev likes to use the analogy of a volcano eruption when describing quantum jumps, which are completely unpredictable in the long term, but with correct monitoring, scientists can detect an advance warning of an imminent disaster eruption and act on it before it has occurred.

What ingredients keeps Minev motivated to keep going forward, who already at age 30, has made seminal contributions to his field?

"I like to create things but in order to create, you have to understand something really, really well," he answered. "And to get to that point, you have to have new insights that no one else has had before, so you can now solve problems that were thought to be unsolvable — like predicting a jump before it occurs. And we can use that knowledge to create new technology." At IBM Quantum, Zlatko continues leading in this path, devising new ways toward full realization of quantum computing, helping accelerate the growth of this new industry.

Outside the lab, Minev is the founder and chairman of the nationwide science outreach organization Open Labs, a recipient of the Yale-Jefferson Award for public service, a Member of the Executive Board of the Yale Graduate Alumni, the host of the IBM Qiksit Quantum Seminar Series on YouTube. He is a frequently invited speaker and keynote presenter at prestigious conferences, such as the septennial Quantum Optics conference, CQIQC Bell Prize Toronto, IQC Quantum Innovators, American Physics Society March Meeting, and top universities, such as MIT, Harvard, and Cambridge, and venues, such as the U.S. Department of Energy, and even Springer Nature's headquarters in London.

Manuel Le Gallo-Bourdeau

AI, Deep Learning and Machine Learning are terms that are ubiquitous in our lives today. All of these concepts are based on neuromorphic systems that provide methods of computing inspired by the human brain. However, all of them are also software-based simulations of the neuromorphic system in an actual digital computer. Although not immediately obvious, this is in fact a very inefficient way of computing.

Using decimal arithmetic in computers to perform cognitive tasks such as image and speech recognition is slow and hopelessly energy-inefficient. With the fast-growing amounts of predominantly unstructured data, which is preferably and increasingly addressed by AI, it is doubtful whether we can afford this relatively inefficient approach on a worldwide basis moving forward. Manuel Le Gallo's work in neuromorphic and in-memory computing directly addresses these problems of inefficiency.

Focused and determined, Le Gallo had always been interested in the physics of semiconductor devices, and studied electrical engineering at France's Cycle Ingénieur Polytechnicien from Ecole Polytechnique. His undergraduate work drew him to apply for a Master's degree thesis at IBM Research Europe in Zurich, Switzerland, which focused on the physics of the phase-change memory devices being developed at IBM.

"After a one to two years of work, we began to understand much better the properties of phase-change memory devices. We then thought that they would be very useful for making AI hardware, and that's how the whole project started 5 years ago," said Le Gallo.

As a researcher and staff member of the Neuromorphic and In-Memory Computing group at IBM Research Europe, Le Gallo is exploring ways to use phase-change memory (PCM) devices for non-von-Neumann computing, privy to the fact that it could play a key role in the space of specialized computing substrates for artificial neural networks.

Within this field, Le Gallo's doctoral thesis is arguably one of the foundational works. The first part of the thesis presents detailed physics-based models to describe the various aspects of PCM dynamics. In the second part of the thesis, PCM applications in non-von Neumann computing are introduced. Using a prototype PCM chip comprising 1 million devices, he presents the first large-scale demonstrations to-date of in-memory computing for applications such as compressed sensing and solvers for systems of linear equations.

The thesis resulted in highly influential publications in journals such as Nature Nanotechnology, Nature Electronics, Nature Communications, Advanced Electronic Materials and the IEEE Transactions on Electron Devices, as well as 13 granted patents.

"Making AI hardware requires a broad understanding from low-level device design and fabrication all the way up to high-level deep learning algorithms and applications," said Le Gallo. "Getting a good understanding of this broad range of topics to be able to communicate and interact with experts across the different fields in our team has been one of the most challenging tasks for me in this project."

Looking ahead, Le Gallo isn't planning to slow down and the former professional drummer would like to focus on sustainability within AI technologies.

"Training a state-of-the-art AI model today has the same carbon footprint as five average American cars over their lifetime. This is not sustainable given that the developments in AI technologies are growing at an exponential pace," he said.