Where a ship goes after Singapore depends on where it came from. A higher-order network keeps that memory by splitting Singapore into Singapore|Tokyo and Singapore|Shanghai. Redrawn from Fig. 1 of the HON paper.
What I do now
I lead the Data Strategies Group at Citadel, the firm's central research team for alternative data and machine learning. We turn trillion-scale alternative data into signals and forecasts for investment teams across equities, fixed income, commodities and quant strategies.
I joined Citadel as a data scientist in 2017, set the firm's standards for building alternative data panels and correcting their biases, became Quantitative Data Research Lead in 2021, and Head of DSG in 2025. Lately I have been building multi-agent AI systems that automate parts of quantitative data research.
During the pandemic, when alternative data became the fastest read on the real economy, I built the Alternative Data Observatory, which more than 30% of Citadel's investment professionals used during that period. It laid the foundation for how the firm evaluates alternative data, turns it into features, and builds data-driven forecasts.
We are hiring researchers and engineers who like to find signals from real-world, real-time, large and messy data.
who still loses sleep over a new finding, whether it is a higher-order pattern in global shipping or a signal hiding in trillion-row alternative data.
A creative programmer
who loves building tools: HON and HoNVis in graduate school, the Alternative Data Observatory at Citadel, and at home a data-driven system that automates the whole house.
A critical thinker
who leads a research team by asking the right questions, from how a data panel is built and bias-corrected to which research is worth doing at all.
My desk at Notre Dame, June 2014. The Kindle is open to Hackers & Painters.
Selected research
HoNVis, exploring the global shipping network as a higher-order network.
Network science, 2014 to 2022
Higher-order networks
Most network models assume the next step depends only on where you are now. In shipping, web browsing and other sequence data, it often depends on several steps back. HON finds how much history matters and builds it into the network, so clustering, ranking and random walks become more accurate without changing those methods.
Retweets and accounts reached in the first hour after a breaking-news tweet, in ten-minute steps.
Finance and social media, 2013 to 2021
Information diffusion and trading
Begun in 2013, this was one of the first studies of how news spreads on Twitter and what that spread does to the market. We tracked retweets of business news in real time: faster spread brought faster trading, mostly by retail investors, and a price rise with narrower bid-ask spreads that reversed the next day. How far a tweet will travel can be predicted from its first ten minutes.
Species flow between ports implied by ship movements, 2005 to 2006.
Data mining for social good, 2013 to 2020
Aquatic invasions through global shipping
Ships carry invasive species between ports in their ballast water. With ecologists and policy experts, we combined shipping, ecological and environmental data into a species flow network, found the clusters of ports that exchange the most risk, and showed where targeted management would pay off.
KDD Workshop on Machine Learning in Finance, program committee every year 2020 to 2026
ECML-PKDD, Area Chair 2021; program committee 2020, 2022
SIAM Data Mining, program committee 2017, 2020 to 2024
Program committee, NeurIPS 2019 Workshop on Robust AI in Financial Services; KDD 2019 Workshop on Anomaly Detection in Finance; BigNet at WWW 2017
Journal reviewer for IEEE TKDE, IEEE TSMC: Systems, IEEE Transactions on Big Data, IEEE TNNLS, Big Data, Knowledge and Information Systems, Social Network Analysis and Mining, Science of the Total Environment, Ecology Letters and Ocean and Coastal Managementsince 2013
Conference reviewer, AAAI and ASONAM
Organizing
All-day hands-on tutorial at ACM KDD, Beyond Graph Mining: Higher-Order Data Analytics for Temporal Network DataLondon, 2018slides
Organizer and session chair, Higher-Order Models in Networks satellite at NetSci Paris, 2018
Teaching and talks
Guest lectures on alternative data, UIUC graduate and undergraduate finance classes 2021, 2022
Guest lecture, Alternative Data for Fixed Income, Notre Dame MBA 2022
Panel, From Students to AI Professionals, RISE AI, Notre Dame 2025
Making and sharing
Room 5023 and our LED sign, Fudan.Accepting the award, 2011.
Room 5023
Back in college at Fudan, the four of us in Room 5023 could not stop taking things apart to see how they worked, and the dorm filled up with electronics we built ourselves. Figuring things out was only half the fun; the other half was showing people. So we filmed 5023大讲堂, a video series that taught other students electronics by building things on camera. It passed a million views, students from all over China wrote in with questions, and national television came to film the dorm. It made us some of China's first generation of online influencers.
Fudan named 5023 its top dormitory in 2010 for our passion for innovation and our spirit of sharing, and in 2011 the dorm was named one of China's ten National College Students of the Year. As head of the dorm, I accepted the award on stage for the four of us.
The same curiosity now goes into my own home. Every new AI capability is something I want to try on real life, so I have built a data-driven automation system that runs much of my home and takes over a little more of daily life every month.
A wall dashboard in the spirit of the Weasley family clock: who is where, the next train, the weather, the day ahead.
A fully personalized climate automation system.
A personalized finance news podcast every morning, which grew out of a Raspberry Pi radio station I built in 2017.
Otherwise
I take photographs, play the flute, and go on long random walks through Wikipedia.
Jian Xu is Head of the Data Strategies Group at Citadel, the firm's central research team for alternative data and machine learning. His team turns trillion-scale alternative data into signals and forecasts for investment teams across equities, fixed income, commodities and quant strategies. Since joining Citadel as a data scientist in 2017, he has set the firm's standards for building alternative data panels and correcting their biases, built the Alternative Data Observatory that more than 30% of Citadel's investment professionals used during the pandemic, and more recently built multi-agent AI systems that automate parts of quantitative data research. He became Quantitative Data Research Lead in 2021 and Head of the group in 2025. He holds a Ph.D. in Computer Science and Engineering from the University of Notre Dame, where he developed higher-order network (HON) models published in Science Advances, KDD and other venues, and a B.S. in Electronics Engineering from Fudan University. He stays connected to the research community as a program committee member of the KDD Workshop on Machine Learning in Finance.
