Women in Data Science Worldwide

Data science and genetics are closely linked and have been for some time. But now, data science is playing an even larger role in genetics, a trend that is prompting researchers to look hard at their ethical responsibilities, says Chiara Sabatti, a professor of biomedical data science and statistics at Stanford University. As is the case in many other fields, geneticists have access to much more data than in the past, and because it is digitized, it can be mined. “Scientists rely on statisticians to mine this data and help them formulate hypotheses,” Sabatti said during an interview recorded for this year’s Women in Data Science podcast at Stanford. Truly understanding and interpreting this data correctly will become increasingly important for the public good as the relationship between accessibility and privacy continues to grow, she noted. Because there is such a wealth of data, there are potentially thousands of hypotheses that could be explored in some cases, an obviously unworkable situation. Data scientists need to determine which of the hypotheses drawn from the data are worth pursuing, says Sabatti. And that means developing new tools “to be able to confidently say to the scientist, ‘these are the hypotheses that you should follow up.’” Sabatti voiced her concerns about the public’s confidence in science. “I am really worried that as scientists we contribute to this by putting forward results that are not as solid as they should be,” she says. “The idea that data speaks by itself is an illusion. It's very important for us to find a way to communicate to the general public what are the challenges of the data analysis.” This is particularly true in genetics, especially in light of increasing fascination with commercial DNA testing, says Sabatti. “I think the public is not aware of all the consequences of putting their data, genetic or not, online and available for mining. I think it's up to us as scientists to try to communicate clearly what it is that we can do with this data and what are the opportunities that come from data sharing,” she says. Beyond genetics, Sabatti cited the need for “algorithmic fairness,” a new concept that seeks to eliminate biases and contribute to a more equitable understanding of data. She is also hopeful for the next generation of statisticians. “I actually look at this field in a very optimistic view. I am amazed by the intelligence and the knowledge of young people coming into it. I cannot keep up with my students or the students in other people's labs. There is a lot of energy, and there's going to be a lot of interesting knowledge that comes out of this investigation,” she says.

Career paths don’t always follow a straight line. Just ask Elena Grewal, whose education culminated in a PhD in education, but who became the head data scientist at Airbnb. In some ways, the leap wasn’t quite as daunting as it might sound. Grewal’s training at Stanford was interdisciplinary, including statistics and econometrics. “Often it’s more about words being different than about skills being different,” Grewal said in an interview recorded for Stanford’s Women in Data Science podcast. At one point, she began to study machine learning and initially thought it was very different from the work she was doing. “Then I started looking at what people do in machine learning, and I was like, ‘Oh, it’s logistic regression, it’s clustering analysis. I do that; we just call it something different,’” Grewal says. Whether it’s called data science or not, many different fields have some kind of quantitative component, and people in those fields who are using quantitative skills may well have the background to become a data scientist, she says. Employees who are not data scientists can learn to understand and use the data their companies collect. Grewal started “data university” at Airbnb, a program that teaches employees at all levels to work with data to do just that. “I don’t want people who have data to be the keepers of knowledge or power, but to share that and to enable every person to be able to think more critically and to be able to make conclusions themselves,” she says. Grewal’s team taught SQL – a standard language used to query databases – to employees and created a database they could use to access company data. Since Airbnb launched data university last year, hundreds of people from other companies have asked Grewal’s team to help them start similar programs. Although undoubtedly successful today, Grewal champions the importance of grit and believing in yourself as a student, as she herself struggled academically when she was younger. In middle school, a “teacher sat down with my parents and told us that I was a really nice kid and that I was going to be fine in life, but I was just never going to be a top student,” she says. She didn’t let it bother her. After working intensively on math with her father, a university professor, Grewal’s grades shot up and she graduated at the top of her class. “I think that was an important early experience: Where you are is not where you can be. It’s important to just work hard, do your best, and see where you can go and not feel limited,” Grewal says.

Consumers know Walmart as a retailing giant that has changed the face of retail in communities across America. But with a data store containing billions of queries and items, it’s also a laboratory for the company’s data scientists and IT professionals who mine and manage it. “We have data scientists embedded in every single team within the company,” says Sonu Durgia, group product manager for search and discovery at Walmart Labs. “Every function at Walmart, from the quality of groceries to the supply chain, has data science embedded in it,” she noted during an interview recorded for the Women in Data Science podcast at Stanford University. Because Walmart’s product catalog is immense, holding the attention of consumers and helping them find what they want to buy is a challenge. “We do not have your attention for the next several hours. We have to show you the right things very, very quickly. So it's a ranking and relevance problem right there, even though it's not coming from a query,” Durgia says. Explaining the insights of data scientists to the business and retail sides of Walmart, people who are not always conversant with technical issues is an important part of her job, she says. Her varied career path has provided her with the expertise to interact successfully with Walmart’s line of business executives. “My engineering degree gives me those tools to really understand the (algorithms) and work with these engineers and very savvy data scientists. My finance background gives me that bird's eye view, understanding what the key things are here,” she says. Because data science is still a male-dominated discipline, finding a role model can be difficult for women in the field. But technology, says Durgia, has enabled new ways for women to find role models. “Back in the day, you would just look at your peer group to find inspiration or even to solve some problems, ask about a concept you didn't get in class. But now YouTube is your teacher. Everything is available,” she says.

Data scientists are involved in a wide array of domains, everything from healthcare to cybersecurity to cosmology. Megan Price and her colleagues at the Human Rights Data Analysis Group (HRDAG), however, are using data science to help bring human rights abusers to justice. The nonpartisan group played a key role in the case of Edgar Fernando García, a 26-year-old engineering student and labor activist who disappeared during Guatemala’s brutal civil war. Price, the executive director of HRDAG, says the investigation took years, but their work led to the conviction of two officers who kidnapped Garcia and the former police chief who bore command responsibility for the crime. “It was one of the most satisfying projects that I’ve worked on,” she says. Price discussed the case in more detail as well as other cases she’s worked on over the years and the role data science played in an interview recorded for the Women in Data Science podcast recorded at Stanford University. For a recent project in Syria, Price’s group used statistical modeling and found information previously unobserved by local groups tracking the damage caused by the war. Similarly, in Mexico, she expects HRDAG to gain a better understanding of in-country violence by building a machine learning model to predict counties with a higher probability of undiscovered graves. Price hopes that in the future human rights and advocacy organizations will have their own in-house data scientists to further combat social injustices around the world, and she believes that data science will continue to play an important role in the field. She advises young people entering the field of data science and social change to learn a programming language, pick an editor and find mentors and cheerleaders to help them along the way.

Fiber optic cables that convey data at high speeds across the globe area is a well-known feature of modern technology. Now, university data scientists have found a unique use for them: monitoring earthquakes.Distributed across Stanford’s telecom infrastructure, the cables have become a seismic array that has already collected data on over 1,000 Bay Area earthquakes, says Eileen Martin, a recent alumnus of Stanford’s Institute for Computational and Mathematical Engineering, now Assistant Professor at Virginia Tech, whose research is focused on seismology. Martin and Nilah Monnier Ioannidis, a postdoctoral scholar concentrating on data science and genomics at Stanford, sat down to discuss the pivotal role of data in their research for the Women in Data Science podcast. Despite coming from different fields, both researchers tout the importance of data in academic research. Genomic sequencing requires vast amounts of data, but privacy concerns mandate important restrictions, Ioannidis says. Consequently, she is collaborating with outside institutions that have already amassed large stores of genomic data to understand its role in the field of genomics. Kaiser Permanente is among those collaborations; the company has already done a large-scale genomics study for Northern California. Martin says that being open with other researchers and sharing ideas is a real plus in the field. Ioannidis echoes these sentiments. While Martin acknowledges the risk that another researcher will use the shared information, she adds, “We’re all busy trying to do our own experiments.” Their advice for students looking to pursue a career in data science within academia: look for new experimental techniques because there will always be an interesting math or computing problem to solve.

“Fail fast” has become something of a mantra in Silicon Valley. But Janet George, the chief data officer of data storage giant Western Digital, has an amendment to that conventional wisdom: “Fail privately.”She suggests that failing privately allows you to open yourself up to discovery and exploration in a safe setting where you are able to take risks. “Carve out time for yourself so you can fail privately. So, you take 20 percent of your time in big initiatives you feel you can really contribute to, but take 20 percent of your time [(where you can])fail privately.” George, who has worked for some of the most important companies in the technology industry, shared this piece of advice, her career trajectory and the role of data science in the storage industry for the Women in Data Science podcast at Stanford University. Although the fear of failure is natural, it should never become a reason to avoid risk, she says. Taking an executive role at a storage company was a risk for George because she knew little about manufacturing before and. “I had to learn deeply about the device physics domain,.” she says. She became familiar with arcane matters like bit counts, failure rates, temperature testing and the impact of voltage on storage cells in order to ensure her success. Now in her fourth year at Western Digital, George continues to notice how much data science comes into play across the spectrum of the company’s business. From manufacturing to security and product development, “every aspect of mathematics, especially linear algebra, plays a very significant role,” she says. “When you think about the computations of scale, when you think about genetic algorithms, its applications, regression-type algorithms, or you even think about neural networks, it’s computationally heavy, it’s mathematically heavy.” Creating a die, essentially a mold, for a new storage device, for example, starts with tens of thousands of possible parameters. Data scientists at the company have to sift through a multitude of mathematical possibilities and discover the 20 or 25 most critical parameters. As the only woman at most executive meetings, George is wielding influence as a lone voice at the table, a skill honed over many years with important risks taken along the way. Her advice for aspiring data scientists: Build relationships and credibility within your organization and lead by example.

When Jennifer Widom began her career in computer science, it was a relatively narrow and specialized field. Three decades later, computer science has become an interdisciplinary field that touches on broad swaths of society and promises solutions to global problems such as healthcare and sustainability, she says. “Computer science used to be a niche. But (it) has become much more broadly used, broadly applicable across all fields. Instead of it just being a narrow study of software and hardware, it's now a lot about what you can use that software and hardware for in other fields,” says Widom. Indeed, learning about the relationships between math, computers and music prompted Widom to make a radical career change. Her undergraduate degree is in music, and she was on a path to become an orchestral trumpet player. But a course focused on computer applications for music was so intriguing she shifted her studies, eventually becoming a computer scientist and the dean of the School of Engineering at Stanford. Increasingly, jobs in industries related to computer science will be broader and encompass the need for data science at its core. “We’ll still need straight-line software engineers, but there will be more jobs for people with additional skills and interests,” Widom said in an interview recorded for the Women in Data Science podcast at Stanford University. That shift may well make the field more attractive to women, she says. Computer science has become so popular that nearly 20 percent of the student body at Stanford is majoring in it, and the university is struggling to keep up with demand, she says. Data science continues to play an important role in its continued evolution as more and more students use data to solve complex problems. But what do those students really want? “Are the students who are coming to computer science coming because they want to learn just the computer science, or are they coming because they want to apply computer science to their other interests? I'm going to venture a guess that the second is true for a lot of those students,”Widom says. If that’s the case, Stanford and other universities will need to shift the computer curriculum to be more reflective of its newly interdisciplinary nature, she says. Widom pioneered the use of MOOCs —massive open online courses —and says teaching them “was one of the most invigorating and exciting things I think I've done in my whole career.” The experience of reaching so many people —her first effort attracted 100,000 students —inspired her to take a sabbatical in which she traveled to under-developed countries offering free short-courses, workshops and roundtables, covering such topics as big data, collaborative problem-solving and women in technology. Her “instructional odyssey” was not only personally gratifying, but it shaped her teaching. “I think, based on my experience with the MOOCs and travel, that the way I could best influence people directly would be to show up and teach them,” she says. “I just really loved reaching people all over the world.”

Be yourself” was just one of the many career tips Caitlin Smallwood shared during a conversation with Stanford professor and Women in Data Science podcast host, Margot Gerritsen. Smallwood, vice president of data science and analytics at Netflix, urges up-and-coming data scientists to explore “the avenues and nooks and crannies” of the discipline and avoid limiting themselves to the most obvious paths. Smallwood is passionate about data-driven content and predicts that deep learning will continue to propel advances in applied data science in the future, specifically in the area of machine translation. It will take some time, she says, but machine translation would allow users to watch a movie or video and understand the subtleties of language and culture at a deeper level through nuances in inflection appropriate for different languages. Smallwood is interested in the ways that data science guides content and helps people “understand regions and cultures around the world through storytelling.” She enjoys the fact that her job allows her to engage and learn as well.“I, myself, have learned so many things from watching different pieces of content. You learn something that’s much more subliminal or that can really impact your empathy when you relate to a character and see the details of how they live their lives in an entirely different culture. And that’s different than reading a news article about a culture,” she says. As to her own future, Smallwood expects to stay at Netflix for a long time. “There are just such massive, new, exciting problems that we’re working on now, and I can’t imagine that changing.”

Attaining tenured status at a major university is often the culmination of an academic’s career; giving it up is unthinkable for most. But after 10 years at UCLA, Jennifer Chayes was offered a job at Microsoft. The offer, she says,“scared me to death,” but she took the job and is now managing director for Microsoft Research in New England, New York and Montreal. “There are brass rings that come along,and they always come along at the most inopportune times,and they look really scary, but I believe that we should grab them when they come along,” Chayes says during a conversation with Stanford’s Margot Gerritsen, Stanford professor and host of the Women in Data Science podcast. Chayes is a big advocate of eliminating biases in search algorithms and believes that data scientists have “the opportunity to build algorithms with fairness, accountability, transparency and ethics, or FATE.” FATE, a group that formed at one of Chayes’ labs, works to address inequity in the field. In one particular instance, the group discovered that certain searches yielded certain results. Searches looking for computer programmers, for example, typically returned results for people with male names. The change Chayes' team implemented in the search algorithm removed that built-in bias. Removing bias from hiring is not only fair, it results in better outcomes, she says. “I think that you’re more likely to ask the right questions if you have been on the wrong side of outcomes. So you’re much more likely to see a lack of fairness or bias as a problem before it happens.” Chayes believes that the fieldof data science is changing and that the increase in underrepresented voices will be critical to the future of the field moving forward.