Target continues to emerge as a data science leader. Whether it’s greater personalization or a nimbler supply chain, our engineers and data scientists help bring to life Target’s next big idea faster and smarter.
The 2020 Grace Hopper Celebration (Sept 29–Oct 2) gives attendees a glimpse into some of the most exciting work to result from our investments in data science and technology—and the women technologists who are propelling Target into the future.
This year, there are two presentations – learn more at jobs.target.com/gracehopper:
DevOps + Data Science: Reduce Time to Data Driven Decisions
Presented by Koel Ghosh, Lead Data Scientist, and Jacob Yunker, Lead Engineer and Software Engineering Coach
Shop the Look: Fashion Compatibility Prediction for Complementary Product Recommendation
Presented by Luisa Polania Cabrera, Principal AI Scientist, and Yiran Li, Lead Data Scientist
Read on for a glimpse at the fascinating content from the Dev Ops + Data Science presentation—plus insights on why this work matters, how it both delights and challenges those who choose it, and how Target’s commitment to diverse thought partnerships gives us an advantage from the inside out.
Can you give us a “sneak peek” of your presentation at the 2020 Grace Hopper Celebration?
Koel Ghosh: Our presentation is based on work we did in December 2018 to launch an interactive dashboard on the Target Application Platform. The solution delivered on many aspects, including accessibility, scalability and customizability, and enhanced my productivity as a data scientist. Coaches taught team members about it and it was well adopted. And we didn’t realize when we submitted the work to the Grace Hopper Celebration that the solution stack also increases productivity in remote work environments, which is so very valuable and necessary during the ongoing pandemic.
Jacob Yunker: This joint effort arose from the need to create a custom data visualization that was scalable and easily shareable with end-users—all while reducing as much engineering maintenance as possible. We have data viz solutions at Target like Domo, but they don’t always do what a data scientist needs, or they don’t do it easily. Many data scientists can make their own app, but then there are issues around where to host it for business partners to easily access it, as well as how much long-term maintenance is involved with owning a custom app. What started as an experiment has since been repeated many times by other data scientists. In fact, the solution is now a training that my team hosts multiple times a year. Having worked for years in data science and now being a Software Engineering Coach, I love helping bridge the two worlds of data science and software engineering.
Can you share a fun fact about Target’s data science team?
Koel Ghosh: It includes folks from many disciplinary backgrounds—computer scientists, physicists, mathematicians, statisticians, etc. In fact, I am an economist by training, and one of the few social scientists here.
Jacob Yunker: It’s such an innovative space, with so many talented folks trying out new ways to bring data into business decisions. As I coach, I continue to feel so fortunate to work with this organization. It’s so energizing to help data scientists bring amazing ideas to life using patterns that are as robust, scalable and automated as possible. My personal belief is that a data scientist’s time is best spent doing data science. Any necessary engineering should support this effort, not hinder it.
Why is data science important at Target?
Koel Ghosh: Data Science enables Target’s purpose to come alive more easily and efficiently. We get to leverage data, algorithms and technology to make life easier for both guests and team members. We save them time so that they can reach out for more meaning and more joy.
Can you describe a time when Target incorporated diverse perspectives to influence a project outcome?
Jacob Yunker: I see many examples of this, and my partnership with Koel over the years is a great one. We have different backgrounds, strengths and perspectives. By partnering on that original experiment (the basis of our presentation), we have realized so much value, not only from that product, but in many others based on the same solution. I don’t think this would have been possible without two people with different experiences working together to try something new.
What do you like most about data science? What do you find most challenging?
Koel Ghosh: It’s a wonderful spectrum of varied activities involving data coming from many guest-related, business and operational areas. You go from creating a simple report, a dashboard, or an ad hoc analysis, to standing up products that do automatic decision-making at scale. It also involves testing, and contributions of thought from so many disciplines. All of this collectively helps inform better business decision-making and processes that serve our guests. There is no end to the learning, and there are so many ways to approach a solution. I came in with a strong statistics and economics background, and the area that I have leaned into and learned the most about being a more effective data scientist is in computational efficiency, i.e. how to get things to run very quickly. Just keeping up is challenging. The tech stack or methods that serve Data Science evolve so fast. It can be overwhelming if you aren’t disciplined about it.
Jacob Yunker: Based on my time working in Data Science, I think my answer to both questions is the same: The questions we are trying to answer are both enjoyable and challenging. Every data science product I worked on had a very interesting problem to solve, and there were so many approaches to think through and try. Of course, some approaches work better than others, and it’s always a learning experience.
Can you describe some of the leading-edge machine learning AI technologies you’re working with?
Luisa Polania Cabrera (co-presenter, Shop The Look): There are many frameworks and libraries in the AI field, and they are always evolving. My team heavily uses PyTorch and TensorFlow for the development of deep learning models. PyTorch is an open source machine learning library for Python based on the Torch machine learning library that allows for fast and flexible experimentation. It originated in Facebook’s AI research group. Similarly, TensorFlow is an open source machine learning framework that is easy to use and deploy across a variety of platforms. It was created by Google for supporting its research and production objectives.
Finally, what advice would you share with someone considering a career in data science at Target?
Jacob Yunker: Be curious, humble and open to change. Learning is a life-long thing, so you won’t always know the answer, and things can change in the blink of an eye. Challenging yourself to be OK with this is so important for success. Partner with people who complement your skillset and have different perspectives. Diverse teams build better products!