Statistician and data scientist continue to be ranked among the top jobs, most recently by U.S. News & World Report in 2017, which listed statistician as #1 on both its lists of Best Business Jobs and Best STEM Jobs, and Glassdoor in 2018, which had six analytics and data science jobs included in its 50 best jobs in America. With this in mind, we asked leaders in industry to answer the following set of questions to help students and statistics departments better prepare for jobs in the technology industry. ~ Steve Pierson, ASA Director of Science Policy

Demand Intelligence and Experimentation Platform

Olivia Liao leads the pricing team at FLYR to build the next generation of revenue management systems for the airline industry. Previously, she led the data science of Demand Intelligence at Uber, which models and predicts how riders choose different Uber products. She earned her PhD in statistics from Stanford University.

 

Jeremy Gu is vice president of the San Francisco Bay Area Chapter. At Uber, he is a data scientist on the Experimentation Platform and works on the production of experiments. Previously, Gu was an applied scientist at Amazon, working on automated advertising and Amazon Web Services teams. He earned an MS in statistics from the University of Washington and a BS in mathematics and statistics from the University of Minnesota.

Please describe your career experience.

Jeremy: I chose statistics as my major because I loved analyzing data and making predictions. In high school, I always wondered how the big lottery worked. At the University of Minnesota, my first statistics class was STAT 101, taught by Sanford Weisberg and his teaching assistant Christina Knudson. They were very good at putting complex concepts into easily understood stories. So, I found their class fascinating and then was convinced statistician would be a great career for me.

After consulting many classmates, teachers, and people from industry, I decided to pursue a mathematics major, as well. When taking upper-level statistics classes, I realized mathematics, such as advanced linear algebra and real analysis, was critical for me to better understand statistical theories. Further, I took C++ and algorithms classes, because computing was a growing component in statistics curriculum.

For a college student, GPA is important. Many of my friends went to graduate school and the admissions committee valued high undergrad GPA. However, a more important component of building my career vision was talking with people from various backgrounds. I was lucky to meet Lynn Argetsinger at the University of Minnesota when she organized events for the college of liberal arts. I attended the class reunion for the statistics school in 2011 and met many remarkable people who also graduated with a statistics degree. The advice from Stephen Eick, Denise Harbert, and Lynn Lin still influences my thinking and approach.

One of the most valuable pieces of advice I received was to look for a summer internship. I was notified by Gilad Lerman in the mathematics department with an offer to intern at a top research center working with Jiaqi Yang at Schlumberger-Doll Research in Boston. I still remember Lerman worked many hours to help me in the application process between the university and company, and I am still extremely grateful for his help. After securing that first internship, many things in my life got much easier.

I intended to pursue a PhD in statistics and, for that reason, I went to a top statistics master’s program in Seattle designed for later PhD applications. At the University of Washington, the size of the master’s program was tiny, and I cherished the time with my six classmates in the two-year studies. In addition to the problems discussed in the textbook, I had the opportunity to work with real data sets at school. Later, Caren Marzban offered me a research assistant position and we published a paper together.

Tech giants Amazon and Microsoft were growing rapidly in the Seattle area, and I had opportunities to meet people at both companies. As a result, my view of my career changed and I decided to explore job opportunities in the Seattle area. At that time, big data and machine learning were two big directions, and many people started to use data science for jobs working with numbers and models. After getting my MS in statistics from the University of Washington, Amazon hired me in the cloud computing department. One year later, I switched to another team at Amazon focusing on machine learning models on the automated advertising. Now I work on the experimentation platform team at Uber and am still a student who needs to learn and grow.

What do you like about working in the technology industry? What are the challenges?

Jeremy: As a data scientist, I enjoy working in the technology industry because there are many opportunities to apply science techniques to solve challenges. At Amazon, I first worked in the cloud computing business and then learned about many things, including Spark and Hadoop. On the business side, I honed my skills in advertising and marketing. For example, a data scientist can design an online experiment to help choose the right metric to measure an advertising campaign. Moreover, data scientists must know how to build machine learning predictive models in an end-to-end fashion; the entire process includes getting the data, training the model, putting the model into online production, and measuring and improving the performance. Outside the data science domain, you will learn many other interesting problems and meet many talented co-workers from other backgrounds.

As with any other job, there are also challenges with working in the tech industry. First, the tech industry changes very fast and data scientists need to keep up with the latest technologies. For example, after graduating in 2014, I spent another year studying Hadoop Ecosystems and Spark in a certificate program. I found it useful because I didn’t use Hive, Hadoop, or many other new tools on big data platforms at school. Coursera was a great resource for me as a student. I took both free and paid classes in machine learning, big data, digital marketing, and so on. I also recommend attending local meetups to connect with people with similar interests. Sometimes, studying with a study buddy motivates both people.

There is a big difference between learning what’s in textbooks and being able to apply those concepts to industry work. It can take several years for new graduates to adapt to the professional environment. I refer to this time as post-education education. I learned as much during this time as I did at school. For example, I learned how to use several SQL languages on different platforms. Data querying, parameter tuning, outlier detection, and many other techniques have quick and clever solutions once you acquire enough experience. Besides technical knowledge, students entering the workforce will learn how to work with a direct manager, communicate with people from various backgrounds, manage personal income, and many other things that aren’t taught in a classroom.

How is the demand for statisticians in the technology industry? What are the main degrees you consider when looking for candidates with statistical expertise?

Olivia: The demand for statistician remains high, sometimes even when technology companies don’t realize it. In the era of big data, the industry needs statisticians with keen eyes to find insightful business trends from messy, often confounded data, as well as to determine how the company’s products affect behaviors through causal inference methods. Statisticians with skills in high-dimensional predictive models, which sometimes are the core products of the company or team, are also in high demand. Also, even with big data, the business-relevant events data scientists deal with can be rare (e.g., ads click-through, client default on loans, insurance claims, etc.) or the data become sparse when you dig into particular business segments. The ability to handle these analytical challenges is highly desired.

Most hiring managers consider candidates with degrees in statistics, biostatistics, economics, and industrial engineering. However, the exact degree is less important than relevant internships and projects. Also, different companies and teams require a diverse background for data scientists. While A/B experiment and regression modeling are common skills required for all positions, not all teams need experts in machine learning or deep learning. Many look for generalists who have broader skill sets to take on any challenging analytical problems, while some look for experts in causal inference, time-series forecasting, spatial-temporal modeling, etc.

What do you see as the most important statistical skills in the technology industry? What are other important skills necessary for a successful career in this sector?

Olivia: The most important statistical skills we practice daily and are looking for in candidates include experimentation (hypothesis testing), data visualization, regression, and high-dimensional modeling. The first two are often overlooked. However, in our line of work, we are often given vague business problems and need to lay out plausible hypotheses (e.g., factors that drive consumer purchasing decisions). Then we devise experiments—either online, simulated, or observational—to validate those hypotheses. The results can sometimes fundamentally change the direction of the product development and roadmap of an entire team.

To establish a successful career in this sector, data scientists also need strong communication skills. Our work frequently requires cross-functional collaboration with multiple groups. The value of data science will not be fully realized until others understand the value of the work, the complexity in the algorithm, and the exact problem we are trying to solve. “Singular value decomposition” does not convey the problem, and “increase AUC” is not sufficient to get people aligned on the value data scientists bring to the table. Moreover, being able to clearly articulate to your manager any difficulties you are facing can help them resolve your concern most efficiently.

In the industry, data scientists are often asked to tackle vague business problems, rather than textbook-style exercises, and the ability to break down a large problem into smaller pieces is essential. The first step is identifying the underlying business action. For instance, if your project is designing a restaurant rating system, the solutions will differ based on whether the rating is being used merely as a reference for consumers, to rank restaurants in the recommendation system, or to determine prices. From there, a series of sub-questions to address include the type of data needed for the analysis or modeling, any assumptions needed to simplify the problem, potential low-hanging fruit approaches, the success criteria, and how to measure them. To successfully carry out the project, you would also need to identify all the tasks and/or blockers and the best person (not just yourself) to work on each. Together with the team, you decide whether the tasks can be carried out in parallel or need to be sequenced.

Finally, almost all data scientists and statisticians in the tech industry need to code. The task may be as simple as developing an automated metrics monitoring dashboard or as challenging as implementing the model in the production system. In smaller companies, you may even be involved in guiding the design of data warehouse infrastructure.

What advice do you have for students interested in working in the technology industry? Any advice for students in general?

Jeremy: I have two pieces of advice for students looking for opportunities at tech companies.

The first piece is a general one and something many teachers and parents might say: Do your best to study. A solid understanding on statistics, machine learning, and/or big data will be rewarding when working at tech companies.

The second piece of advice is to ask yourself what your strength is and what your interest is. When the two items align, your career path will be rewarding no matter if it is in data science or not. In my view, data science is not a fashion, but a rigorous science after years of hard work from mathematicians, computer scientists, statisticians, economists, and many others. Those people made huge progress in their fields in the past and are still working hard to advance in the new AI/ML technology. I suggest spending a large amount of time thinking about your goals before committing to a data science career.

What advice do you have for statistics and biostatistics departments?

Jeremy: I have received many messages from statistics majors on LinkedIn asking for job advice. I put the messages into the following three buckets:

• Questions about how to choose classes toward a data science career
• Questions about how to apply for data scientist positions
• Questions about how to prepare for job interviews

I believe many of those students didn’t collect enough satisfying answers from the school and therefore asked me instead. I would make the following proposals to the program adviser at a university:

• Encourage students to attend networking events and conferences
• Partner with tech companies to provide internship opportunities to students
• Offer course guides on data science to students looking for a data science job

Invite people from the industry to give frequent seminars or guest lectures to help students understand the industry needs and opportunities

What opportunities for advancement and professional growth exist for data scientists and statisticians in industry, and what advice for young professionals would you have to take advantage of those opportunities?

Olivia: Career growth in the industry can generally be divided into two tracks: individual contributor and manager. This separation is more apparent in large technology companies. Larger companies also have more clearly defined career trajectories and resources for self-development. However, smaller companies like startups usually provide more and faster growth opportunities.

People typically choose the individual contributor track if they are passionate about driving business impact through deep technical expertise and profound domain knowledge but don’t want to spend as much time on people management. In comparison, people with a desire to influence the company through formal leadership roles or team building are ideal candidates for the manager track.

To grab these opportunities, we encourage young professionals to start thinking about their future growth early on in their careers and establish a habit to periodically (e.g., every six months) evaluate their progress. Opportunities don’t just fall into the lap of the hardest-working individual. Proactively discuss opportunities with your manager or mentor and ways you can achieve them. We also encourage young professionals to think beyond technical skills development and constantly practice soft skills—including communication, leadership, and team coordination—to effectively move projects forward. No matter which track you pick, the ability to help others achieve collective goals and deliver the greatest value is critical to a successful career.

Risk Management

Xin Ge is chief risk and analytics officer for Afterpay Touch Group, responsible for risk, data, and analytics initiatives globally. Before Afterpay, Ge was head of risk management at Uber and VP of global seller risk management, resolutions, and protections at PayPal. He was also the founder of PayPal Risk Management Center in China. Ge started his career on risk management at eBay 15 years ago. He holds a BS in computer science from the University of Science and Technology of China and PhD in statistics from The University of North Carolina at Chapel Hill.

Please describe your career experience.

My first job out of school was as a consultant at Insightful Corp. in Seattle. We used S-PLUS to simulate high-frequency trading algorithms for various trading firms. The work was fun and matched well with my PhD thesis, which was related to derivative pricing. It required a good understanding of financial products, as well as strong communication skills to interact with the clients. To create the most efficient equity portfolio, we often needed to solve optimization problems.

I joined eBay in 2002 and spent the next 14 years taking various risk management roles with eBay and PayPal. I started as a statistician at eBay, working on metrics, reporting, and modeling to tackle fraud and abuse in the eBay marketplace, including account takeovers, merchant fraud, unpaid items, and shill bidding. I took on management responsibilities gradually, starting with technical functions at eBay and growing into more general management where I was responsible for end-to-end business results at PayPal. From 2010–2015, I spent five years in Shanghai and built PayPal’s Risk Management Center in China.
 At Uber, I led the global risk management team, working with a cross-functional group of product managers, engineers, and data scientists to tackle unique risk challenges related to payment, account security, and marketplaces.

Two months ago, I joined Afterpay Touch Group as chief risk and analytics officer. Afterpay, which recently joined SP/ASX 200, has had tremendous success in Australia and was named Australia’s Fintech company of the year. I am thrilled to be part of the team, marching on a journey as we expand globally, beginning in the US.

What do you like about working in the technology industry? What are the challenges?

Risk management is a fascinating world, with challenging and engaging work. First, you feel great about what you do as you stand for justice, protecting the company and the needs of good guys. The work is also intellectually stimulating, because you are working against fraudsters—real people who use advanced techniques to cheat. There is a lot of back and forth since every move you make is quickly countered. You also don’t have a choice but to keep up with the latest technology, because your opponents do. It has been a fun journey shifting battlefields from offline to online and mobile and witnessing the evolution of using intelligence such as IP, device, and GPS signals.

How is the demand for statisticians in the technology industry? What are the main degrees you consider when looking for candidates with statistical expertise?

As the technology world has evolved so much in the past 15 years, we have seen an explosion of data needs and a sharp increase in demand for candidates who can make sense out of data and drive business decisions. Our team members usually have advanced quantitative or engineering degrees—statistics, economics, computer science, operations research, and engineering are some of the common ones.

What advice do you have for students interested in working in the technology industry? Any advice for students in general?

The first step of any data analysis is to prepare data. In the real world, data don’t always come in the ideal form. At companies such as eBay, PayPal, and Uber, data often come in an unstructured way and at a super scale. Hence, I think the most basic and important skill for anyone who starts his/her career is the ability to handle complex and large-scale data. Programming languages such as SQL, Python, and R will come in handy for anyone interested in data-related work.

Another important skill is to discover patterns in the data, whether you prefer the term predictive modeling or machine learning. The line has blurred quite a bit between traditional statisticians and computer scientists, and we have a lot more computer scientists in this field than 10 years ago. The following is a generalization and certainly not necessarily true for everyone, but I have observed that statisticians tend to have a deeper connection to the business problem, trying to link the patterns with common sense so they can interpret the results. Computer scientists, on the other hand, are usually stronger in manipulating and studying large-scale, complex data and implementing models in a production environment faster. Both computer science and statistics are important, and a combination of both skills would make someone valuable.

What advice do you have for statistics and biostatistics departments?

To prepare for industry work, it would be beneficial to add coursework and/or experience for students to process real-world data and solve real business problems. I would also encourage students to look for summer internships, because they are a valuable experience that will not only help students develop technical skills, but also identify how they can fit into the actual work environment.