Written by: Matthew Busekroos | Produced by: Nate Caldwell
Originally from northeast China, Zhizhen Zhong completed his undergraduate and PhD studies in electronic engineering, in particular optical communication and networking systems, at Tsinghua University in Beijing. Zhong wanted to come to MIT’s Computer Science and Artificial Intelligence Laboratory for his postdoctoral training. According to Zhong, this is the place where he could pursue interdisciplinary research right at the interface between electrical engineering and computer science.
He said, most importantly, the MIT motto “mind and hand” inspires him to not only think critically, but also get his hands dirty by building real prototypes and testing it in the real world.
“The CSAIL is such an inspiring place because you are surrounded by a group of brilliant people, from faculty to postdocs and students,” Zhong said. “One of the most important things that I learn here at CSAIL is to really focus on the big problems in the community and be brave enough to take it– ‘dream big and work hard.’”
Zhong said curiosity drives his research. He added that he is always enthusiastic about exploring what can be done with optical technologies to boost the performance of today’s computer systems. “By engineering the unique properties of light and its fundamental particles photons, we could unlock so many new functionalities otherwise not possible with electronic systems”, said Zhong.
Zhong is currently working to realize the vision of how to use light to perform computation at a much faster speed with much less energy, and rearchitect the data movement pathways through the co-design between emerging optical technologies and existing computer systems stacks, contributing to a future where data flows across next-generation computer systems with minimal or zero bottlenecks. In particular, in the past three years, he is the lead on a project known as “Lightning” that brings the best of light and electrons together to accelerate data - intensive applications like machine learning and database lookups.
“I lead a team of students to first propose a novel abstraction that allows novel optical computing hardware to be reconfigurable for different computation tasks, and work seamlessly with today’s electronic computing infrastructure like digital memory and ethernet network interface, then implement the first photonic computing prototype to serve real-time machine-learning inference requests,” Zhong said. “Our experimental prototype also features a record-breaking end-to-end photonic computing frequency at 4.055 GHz.”
This is an interdisciplinary project, and the group’s results were published at both top scientific journals such as Science and top computer science conferences like ACM SIGCOMM.
“The photonic computing system that I am working on has the promise of accelerating data-intensive applications like machine learning by tens to hundreds of times, while reducing energy consumption also by hundreds of times compared with today’s GPU servers,” he said. “With the end of Moore’s law and the rise of generative AI, the world needs a new computer system paradigm to sustain the computing need without exhausting the energy on the planet. I think my research in photonic computing paves the way towards the goal of sustainable computing for the explosive AI workloads.”
“Decades ago, computers were built to run with electricity and connected by electric wires,” Zhong said. “In the past several decades, we have witnessed the communication between computers are largely replaced by optical fibers because optical communication gives us higher bandwidth and less energy cost. Moving forward, I think this trend will sustain – I am curious about continuing pushing the limit and replacing even the computing units with light to gain more even higher speed and less energy consumption.”
Additionally, Zhong also worked on optical innovations in the global Internet. His first project after joining CSAIL was on reconfigurable fiber optical network in the wide - area networks. He built a system called ARROW that restores the lost capacity of fiber cut failures.
Zhong said in the long term he would like to pursue a career in academia because it allows him academic freedom to explore the endless frontiers of computer systems, propose bold ideas and build small yet convincing prototypes to verify.
“I enjoy academia because I love teaching and working with students,” he said. “I like the process of exploring the unknown, then consolidate what my team has found during the exploration and put them into course materials.”
Zhong gives guest lectures about his research findings at computer science courses here at MIT. Working with his colleagues at RLE , Zhizhen recently developed and co-instructed a new course, 6.S918 Optical Computing in the Era of AI, to cover the state-of-the-art topic of optical computing to disseminate the knowledge about this emerging discipline.
“At the same time, I do realize that nowadays, especially in the field of computer science, many problems lies in industry scale where companies need to deploy actual systems and serve billions of users worldwide,” Zhong said. “Therefore, I will also keep my research efforts closely tied to industry and seek academia - industry partnerships . For example, one of my previous research projects on reconfigurable optical wide - area networks that can re store from fiber cut failures ca me from a deep collaboration with Meta (Facebook at that time). I proposed an academic idea to solve the problem of fiber cuts, and built a system at Facebook scale to verify tha t the idea actually works in the real world.” Another example is that, to democratize photonic computing and make it accessible to a broader audience, Zhizhen is now working together with optical equipment vendor, Thorlabs, to develop open - source software and hardware for the first publicly available photonic computing educational kit.
For more information on Zhizhen Zhong, check out his webpage: http://people.csail.mit.edu/zhizhenzhong.