Research Scientist Fan Du explores how we can learn from complex data by transforming it into insightful visualizations. His enthusiasm for inventing new technologies took center stage in the two Sneaks he presented at Adobe Summit conferences—Journey Genius in 2019 and Segment Tuner in 2021. His work has contributed to the research community and to Adobe Experience products in numerous other ways since he joined Adobe Research as a full-time researcher in 2018. Here he shares the inspiration behind his innovative work and career path.
What kind of research do you do?
I work on human computer interaction. We focus not just on the computer side, but on the human side—for example, how can we make machine learning models usable by humans, to solve real-world problems?
Just having a good dataset or a powerful machine learning model is not enough. You need to help people interpret the data and understand the model, and data visualization can do that. Data visualization is a way to bridge humans and data. The goal is to help people consume data in a way that makes sense for them, so they are able to use that data to make decisions and drive changes.
This approach has a long history: In the old days, people found out that they could draw a map of a city and diagnose the source of disease, just by looking at where the people would get their drinking water. Now we are connecting data visualization with other topics, such as big data, AI, and machine learning.
How did you get interested in this field?
As an undergraduate, I did a project that I didn’t realize was about data visualization! It was about how to build a dashboard from data. I looked at pictures, and I saw that one used a treemap visualization to communicate about the stock market. I realized it was so powerful to look at a picture rather than only reading a news article. (Treemaps are currently used as a visualization tool in Adobe Analytics.)
At that point, I got into data visualization. I loved how you can show data in a beautiful picture. But it’s not just about producing beautiful pictures—it’s about making them understandable and insightful for people.
How can this work help people?
In general, we want people to be able to understand and trust data, and to be able to make decisions from it.
We have more than one audience for this work. First, there’s the general public. They want charts they can understand. We try to keep things simple so the data is clear. We can highlight the insight and use context to explain that insight. Another audience is marketers, analysts, and data scientists. They ask for things like, “show me the output of a million predictions from my model,” and “help me find where the bug is in this system.”
Could you tell us about the two technology Sneaks you presented at Adobe Summit?
My first Sneak was Journey Genius. We wanted to address marketers as our audience and explain the output of an AI model, which can predict what their customers will do next. This project visualized this complex information for marketers. It was covered in Forbes, and it will now contribute to an Adobe Experience product, Adobe Journey Optimizer.
In my second Sneak project, we dug into the data itself. Segment Tuner focused on data quality. Usually people think first about a use case. But before that, you need to take a look at the quality of the data you are relying on before you run use cases. If that data is not good, your work won’t be reliable. So we worked on data quality issues that could be repaired before they go downstream.
What sparks your curiosity in your research field today?
First, working at Adobe Research keeps me super interested in this field. I can work with real users and marketers in a high-stakes area for them, building tools and helping real businesses.
Second, the growth of AI and machine learning is giving us many new models and capabilities that were never possible before. I want to explore how to combine those new possibilities and make them useful.
Third, we’ve seen the power of data during the pandemic. Before, many people wouldn’t take the time to read data charts, but now, they are trying to build up their literacy for reading data. This shows how important it is to make data visualization accessible.
Could you tell us about your career path? What led you to Adobe Research?
I did an internship here unintentionally. During my PhD at the University of Maryland, I was talking to my professor, who had contact with Adobe Research. I didn’t realize at the time that Adobe had this whole business in digital experiences. Then I saw that there were so many business use cases related to my research. I wanted to do research work that was more applied, and I hoped to help people with my technologies. Adobe Research was a great fit.
As a result of my internship, we published an academic publication in ACM Transactions on Interactive Intelligent Systems in 2019, and it won a best paper award. The paper was focused on how people can find others similar to them and learn from them. For example, in the medical field, doctors could find similar patients to theirs and learn what diagnosis and treatment they received, to help them work more precisely. We have also done a case study at Adobe, helping marketers find similar customers and understand them better.
What is it like working at Adobe Research?
The collaboration here is just great. There are so many talented people. I talk to different teams all the time. Every day becomes a new day, with new initiatives, and that’s the kind of environment that I wanted for my career.
If you are interested in both research and business impact, Adobe Research is a perfect place. Take a look at our demos, sneaks, papers—you will find a great balance.
Based on an interview with Meredith Alexander Kunz