Our November meeting featured a panel on Data Science, moderated by Board Member Elaine Cheng and featuring four local women working in the field:
- Jade Kim is a Senior Marketing Science Analyst at Analytic Partners, an industry leader in providing data-driven marketing consulting for global fortune 500 clients. Based in the Charlottesville office, Jade focuses on leveraging advanced analytics to derive actionable marketing insights from a world of big data.
- Taryn Price has an extensive background in machine learning, modeling, simulation, and statistical analysis. She has data science experience working on projects in many different industries including health care, energy, and national security. Taryn’s current work at CCRi focuses on uncovering relationships between entities by combining many data sources into a single model-space.
- Monica Rajendiran is a recent graduate of the UVa DataScience Institutes’s MS in Data Science program. For her graduate capstone, she worked with the UVa Health System to design and build machine learning models to predict patient 30-day revisit risk.
- Kim Scott is a co-founder and the Vice President of DataScience at Astraea, a Charlottesville startup dedicated to solving complex problems about the Earth. She is passionate about wrangling data and using novel machine learning methods to help steer organizations towards better decision making.
Question: How did you get started in this field? What attracted you to it?
Ms. Scott, who has a PhD in Astronomy and held two post-docs in the field (the latter in Charlottesville, at NRAO), began to question her initial career choice when confronted with the slow pace of learning that is a given of radio astronomy research. Seeking advice from other ex-astronomers, she learned that data science had a lot of overlap with her existing skill set. After networking her way into a position at Elder Research, she has never looked back.
Ms. Price has an engineering background, and when she ended up teaching seventh-grade math (a story for another day), she found herself collecting data on her students. Eventually, she came across systems engineering, received an MA in that field, and from there discovered data science. Coincidently, her first job in the field was also at Elder Research.
Ms. Kim did her undergrad at the McIntire School of Commerce at UVa where she focused on marketing and IT. Although she initially leaned toward the creative side of the field, once she took a business analytics class, she discovered a more technical interest. After an internship at Analytic Partners, she began full-time work there.
Ms. Rajendiran pursued Biology and Computer Science as an undergrad. She worked as a systems analyst but missed the creativity of some of the other tech she’d done in school. After hearing about it from a friend, she discovered and subsequently completed the Data Science MA program at UVa.
Question: Data science and its insights are often referred to as “magical”. How do you describe it?
Certainly data science is perceived that way, says Ms. Kim. But it could be defined as simply as taking data and deriving meaning from it. It’s not just the data inquiry, however, adds Ms. Price. Data science must contain a certain evaluation component.
Ms. Rajendiran agrees, noting that the methods are designed to be empirical (like science) and that the results cannot be taken simply at face value, but instead subjected to rigorous evaluation. Ms. Kim adds that many times the public has misperceptions of data science, first because bad news is better news (you hear about failures) and second because many articles and blogs emphasize attractive graphs at the expense of a discussion of the analysis.
Question: Can you give us some examples of data science projects?
- Predict hospital readmissions in a 30-day discharge period.
- Use computer visioning to do object detection and video captioning.
- Tell a business how to most effectively allocate its significant marketing resources.
- Clean up very problematic data on oil well sensors in order to predict which ones will freeze
Question: What happens after the data is analyzed? How does it turn into action?
Apparently, data science is only as useful as it convincing; even when the client has requested the analysis, it can be difficult to translate that willingness to enquire into a willingness to change based on the findings. This is why, as Ms. Price notes, how, when, and to whom one presents the data is key. Getting buy-in early on, and keeping communication open with the people the potential changes will impact most directly is essential.
Sometimes, as Ms. Scott told us, special knowledge is required to unlock the significance of data science findings. After finding clusters in the log files of all the users of a 3D software program, the only folks who could make meaningful sense of these clusters were the technical staff, who could see right away that the clusters revealed use patterns that correlated with different types of design projects.
Question: What about the trend of data science excites you? What frustrates you?
Our panelists see very clearly the possibilities and limitations of this new field. Ms. Scott is excited about artificial intelligence, a phenomenon she calls the next electricity in terms of its potential revolutionary impact on society. Ms. Kim is excited by the very newness of the field, and its growth potential. But along with that excitement, there are disappointments. Ms. Scott laments how the buzz can distract people from the real problems that data science could be used to resolve. Ms. Kim is frustrated by how the demand for data science does not yet match the educational/cultural capacity to produce quality data in more contexts.
Ms. Rajendiran, who is also excited about AI and machine learning, is frustrated by lack of transparency around algorithms—even if those algorithms are used to do ‘great’ things. She would would like to see people be more skeptical of this lack of transparency—and you don’t need to be a data scientist to demand a greater degree of openness into these tools that are increasingly influencing our lives. As we get closer to having computers learn to do hard things (as Ms. Price notes), we are going to need to get better at asking the hard questions about the data science used to do them.