Dr. Sonika Johri is a Senior Quantum Applications Researcher at IonQ, a company commercializing trapped ion quantum computers. Her research centers around translating the rapidly expanding capabilities of quantum hardware into measurable advantages for end users of quantum computing. She leads IonQ’s efforts in developing applications for current and near future generations of quantum computers. She has co-authored several publications involving cutting-edge quantum algorithm demonstrations in the areas of generative and discriminative machine learning, condensed matter physics, quantum chemistry, and optimization, across a variety of quantum hardware platforms. Her work emphasizes the co-design of quantum algorithms, software, control, and hardware in order to come up with practical and scalable solutions that are aimed at making quantum technologies highly impactful as soon as possible. Prior to IonQ, she worked as a quantum algorithms researcher at Intel Corporation. She has a PhD in theoretical condensed matter physics from Princeton University focusing on strongly-correlated phenomena including the fractional quantum Hall effect and many-body localization.
IonQ is a startup that is commercializing trapped ion quantum computers. Its quantum computers are accessible over the cloud, including AWS, Azure and Google Cloud, and support all major quantum software frameworks. In this talk, I will give an overview of recent results that illustrate how IonQ systems have been applied to solving diverse problems such as image recognition and generation, multivariate data generation, speedup of Monte Carlo methods, optimization, condensed matter physics, and quantum chemistry. I will then focus on three papers that show how hardware advances can combine with innovative algorithms to enable quantum advantage. The first concerns generative learning of joint probability distribution functions via a family of multivariate distributions with uniform marginals called copulas. We design variational quantum circuits to model copulas and use communication and computational complexity results to argue that they have exponential advantage in expressivity over classical models [1]. We demonstrate generative learning based on these circuits on IonQ quantum computers and show that our results outperform those obtained through equivalent classical generative learning. Second, we will discuss results from [2] showing that a quantum Nearest Centroid classification algorithm, using techniques for efficiently loading classical data into quantum states and performing distance estimations, matched the accuracy of classical nearest centroid classifiers on the MNIST handwritten digits dataset when implemented on IonQ quantum computers. The third example will focus on the experimental demonstration of algorithms that reduce the resource requirement for achieving quantum speedup through quantum amplitude estimation. I will conclude by discussing how application-oriented benchmarking can measure and stimulate progress towards commercial applications of quantum computers.
References:
[1] arXiv:2109.06315
[2] npj Quantum Inf 7, 122 (2021)