Jelani Nelson

Jelani Nelson came to Berkeley in 2019 from Harvard University. He is a professor in the Department of Electrical Engineering and Computer Sciences. He was an Alfred P. Sloan Research Fellow and received a 2017 Presidential Early Career Award for Scientists and Engineers.

“I find it fulfilling to work on algorithmic problems that are both practically relevant and simultaneously mathematically beautiful.”

Research Focus: Sketching and streaming algorithms for big data, and dimensionality-reduction techniques for high-dimensional data.

How is this research applied in the world?

Trend detection (answering questions like "what popular keywords are people searching for?" or "what's a popular item that's been purchased over the last 24 hours?"), anomaly detection, minimizing communication in distributed systems, speeding up machine learning algorithms for highly featurized data, and federated learning (machine learning when data is spread over a large number of clients, e.g., users holding their iPhones), to name a few.

Why is this area important to you?

Decision-making needs knowledge, which is extracted from data, and this is a truth that is universal across many domains, e.g., when creating government policy or when making business decisions. Often these decisions need to be made in real time, or based on data spread over many clients, and thus time- and memory-efficient sketching and streaming algorithms become necessary. Aside from practical importance, modern research in this area has both used and developed much beautiful mathematics, ranging from information theory to complex analysis to spectral graph theory. I thus find it fulfilling to work on algorithmic problems that are both practically relevant and simultaneously mathematically beautiful.

Is it true you are one of the fastest typers in the world? How many WPM? 

My "Best Race" WPM is 232, though for longer stretches of time 150-160 is more typical. You can see my record on TyperA, in which I maintained ~170 WPM over two minutes typing sentences (set the record almost 10 years ago and am still ranked 4th there). All of these WPM measures take errors into account via some penalty system.