LIDS alumni are a distinguished group. They include winners of top awards in the field, authors of classic texts — many of which are still being used today — and, outside of academia, a rich tradition of starting innovative and successful companies.

LIDS graduate student Qingkai Liang hopes to carry on this tradition in the months to come, turning his training and ideas towards a blockchain startup, called Celer Network, he’s co-founding with three friends.

But let’s rewind to his five years at LIDS, and how his time here feeds into his path ahead.

Born in 1990 in the Chinese province of Anhui, not far west of the bustling metropolis of Shanghai, Qingkai was part of the world’s first generation to grow up with regular access to the internet. He found himself fascinated not only by online gaming platforms like World of Warcraft, but also with the communications and networking technologies underlying them, which allowed players to connect with each other and complete challenges together in real-time.

“My generation has experienced a lot of innovations in the networking field, and it's an area that has really changed people's lives,” he said. Observing those impacts led him to study information engineering at the renowned Shanghai Jiaotong University, where he graduated top of his class.

He was drawn to MIT, and to LIDS specifically, to continue his studies because of Professor Eytan Modiano’s work on communications networks. He has been part of Professor Modiano’s group since 2013, where he’s studied the fundamental performance limits of networks, including how much traffic a given network can support, and how to design efficient algorithms to reduce network latency – how long a bit of data takes to travel from one point in the network to another — in order to maximize both speed and accuracy of information transmission.

It’s the kind of foundational research that helps Qingkai and his colleagues better understand why the best or most commonly-used algorithms are robust, too. For example, the suite of communications protocols known as TCP/IP (transmission control protocol/internet protocol), which ensures packets of data arrive at their destination in the correct order, has been in use for close to 40 years. However, Qingkai explained, these protocols were developed and first understood based on theoretical models in which network traffic arrivals, transmission rates, and other elements were stochastic — occurring in random patterns.

While these theoretical models work well for a substantial portion of network traffic, there are certain real-world conditions, such as hackers’ attacks on a network, in which traffic is not stochastic, but rather follows a known and predictable pattern. Qingkai studies such adversarial phenomena. For instance, in what’s called a distributed denial of service (DDoS) attack, hackers direct a coordinated flood of requests, originating from many different sources, at a single host connected to the internet. Those requests overload a website so that legitimate users can’t get a response from it. In order to manage a network that may be hacked in such a defined way, both in terms of prevention and recovery, it is useful for theory to encompass stochastic and non-stochastic scenarios. Qingkai’s work centers on this idea — it aims to expand existing theoretical models to accommodate this kind of adversarial behavior, making the models more practical and more generalizable, ultimately optimizing network performance in different types of complex environments.

Those extended models and analytical tools can in turn be used to help build new, more robust network operation algorithms – and help determine when efforts to do this are worthwhile. “I think [the new models] will give us a sense of the scenarios we can tackle, so we don't have to design algorithms that target some impossible scenario,” Qingkai said.

But 2018 is his last year at LIDS, and he is expected to graduate this summer. His next adventure: moving to the Bay Area to work at that startup.

The company focuses on secure, scalable, blockchain-based solutions for cryptocurrency payments. In a sense, the solutions his firm is developing bear certain similarities to communications networks. A communications network moves bytes and bits of data, and its capacity is its transmission rate – how many packets of data can be transmitted per second. Meanwhile, a blockchain database, shared by all participating nodes or computers, is used to enable frictionless value transfers (e.g., money transfers). Its capacity is the amount of money that is deposited, which fluctuates based on the payments made. Due to the similarities, one of the key algorithms his startup developed is inspired by the BackPressure routing algorithm that was originally proposed for communication networks.

Qingkai feels his work at LIDS, especially on adversarial networks, has equipped him well to develop robust and secure algorithms for such scenarios. “The type of research LIDS is doing is more theoretically oriented, but it’s equipped me with logical thinking skills that readily apply, even though this application is in a totally different field.”

Beyond work, Qingkai also enjoys skiing (“In New England it’s the only thing to do during long winters.”), playing tennis, and learning about wine through wine-certification courses (and of course, practical tasting).

In the long term, he hopes his research, whether on adversarial networks or in the blockchain business, will have widespread impact, just as communications technologies have touched lives globally. “Nowadays, we’re still using some of the tools that were developed at LIDS in the 1970s and 80s. It's really cool to imagine how our work right now might influence the future, 20 or 30 years later.”