Harnessing the Power of Quantum

Early in a career that’s gotten underway as the potential for quantum computing has begun to accelerate, Kate Smith is focused on leading both academic researchers and companies toward a future in which the vast power of quantum becomes scalable for an exciting array of uses.

An assistant professor of computer science, Smith arrived at Northwestern two years ago after managing the software engineering team at Infleqtion responsible for maintaining the company’s physics-aware quantum compiler Superstaq, where she directed research and development for projects involving optimized compilation, error mitigation and simulated quantum programs. Before that, she had been a Chicago Quantum Exchange (CQE)/IBM postdoctoral scholar at University of Chicago, and she received a PhD in electrical engineering from Southern Methodist University.

In the quantum architecture lab at Northwestern, called QuantA, Smith and her group undertake research into the architecture and software stack for quantum computing systems. “The software is ultimately what brings the users to the system,” she says. “We interact with our phones, with our computers, through all these interfaces that are human-intuitive. And then, behind the scenes, all the software translates down our inputs and our requests into executables that the hardware understands. So that is really where my expertise lies.”

QuantA delves into how to make quantum computing architecture and models more reliable and scalable through context-aware optimization. “That context is provided through the hardware, and the algorithms themselves,” she says. “We need to leverage different optimizations to make quantum programs have the best chance for success.”

During Smith’s graduate school years at SMU in the 2010s, quantum computers were just starting to come onto the scene, and by chance she happened to take a quantum computing class. She was fascinated by the fact that when running different circuits during what was the noisy intermediate-scale quantum computing (NISQ) era, they didn’t work particularly well, and the hardware had a lot of constraints. “When you were trying to map a circuit to the processor, it was very tedious. You needed to hand-optimize circuits in order to make something runnable,” she says. “The software space for quantum systems has grown a lot since the time I was in grad school.”

Smith’s research—and that of others in the space—has been motivated by the fact that while hardware continues to improve, doing so is very labor-intensive and only progresses at a certain rate. “There are a lot of steps involved with developing hardware, in terms of designing a processor, building said processor, testing said processor,” she says. “And there are ways that we can accelerate the improvements of these systems through software.”

In thinking through how to optimize circuits, Smith has worked from graduate school until now to develop automated tools rather than slowly hand-optimizing everything. “Software is easy to develop, easy to test, easy to deploy, easy to change,” she says. “So we’re able to build software that’s able to, with existing hardware, get the most out of that hardware. And with the co-design or software and hardware, we can, as we’ve seen in classical systems, accelerate the rate at which systems scale.”

As a CQE post-doc, Smith worked closely with IBM and focused on the company’s superconducting transmon-based hardware. She’s currently involved in a project through Northwestern—along with Prem Kumar, professor of electrical and computer engineering, and Jakub Szefer, associate professor of electrical and computer engineering—helping Allstate Insurance explore quantum applications.

“We’re at an inflection point in terms of hardware being deployed that is actually starting to do things reflecting what we would expect a quantum computing processor to be able to do, in terms of demonstrating superposition, entanglement, early error correction, and what-not,” she says. “In this really fast-changing world of quantum computing, we’re just trying to see what the capability of quantum computers are, specifically for the task of quantum machine-learning-based algorithms and applications.”

Smith believes these sorts of academic-industry collaborations and lines of communication will continue to grow. “Because there are a lot of similar problems that are being worked on, and a lot of open-ended questions; for instance, we still don’t know what the standard quantum processor technology will be,” she says. “There’s a variety of physical qubits that are out there.”

Otherwise, within QuantA, Smith and her colleagues have a few research thrusts focused on quantum-related challenges during runtime:

• Quantum error management: This includes low-level quantum error suppression, which involves optimized compilation manipulating circuits to have the most concise, precise implementation when translating from a logical, conceptual circuit to an executable. “We want to make sure that no syntax is lost going from the program abstraction to what’s actually realizable on hardware,” she says.
• Quantum error detection and mitigation: This is emerging as an intermediate strategy for improving reliability and utility of quantum systems that requires some co-processing from classical computing systems, and quantum error detection can provide a “finger on the pulse” of conditions during run-time. “You can detect errors and take actions based on that,” she says. “We’ve looked into different ways that we can improve how we target and map to hardware based on error detection. Error detection also can be used to implement error mitigation, so you can filter out errored outcomes.”
• Quantum error correction: This requires “the highest overhead” in terms of needing classical co-processing, to not only detect but also correct errors with one’s decoder. “We’re especially excited about a lot of the promise that heterogeneous architectures have, she says, meaning “systems that are comprised of different qubit technology, or even qubits that are of the same technology that have different properties. The use of modular systems that have many components working together has the potential to amplify the benefits of individual pieces of hardware, making the highest-quality quantum systems.”

Smith and her colleagues also are excited about the advent of qudits, units of quantum information with more than two basis states (as opposed to qubits, analogous to the bits of classical computing their values of zeros and ones). “You have more logic levels available to you, and that’s something that can allow for a lot of advantages, in terms of a substantial amount of information that you’re able to examine during processing,” she says.

During the next five years or so, Smith expects the potential of quantum error detection to blossom, providing ever more intelligence about and capability for quantum systems. “Quantum error correction, of course, is something that will be necessary for utility-scale quantum computing,” she says. “Coming up with better compilation techniques for quantum error correction and logical qubits, that’s something that we’re really interested in. Those are the key areas we’re excited about.”