Photo by Glen Noble via Unsplash
On Tuesday, we hosted a live Ask Me Anything session on our Community Slack channel featuring Metis Sr. Data Scientist Kimberly Fessel, who took questions about her transition from academia to data science. Kimberly holds a Ph.D. in applied mathematics from Rensselaer Polytechnic Institute and completed a postdoctoral fellowship in math biology at the Ohio State University. She now teaches the bootcamp and says that her enthusiasm for teaching comes from these days as an academic, but along the way, she realized that academia wasn't her long-term passion. She wanted to transition to data science and work with data storytelling, using the power of data visualizations to challenge pre-conceived notions.
Before joining Metis, Kimberly was working at MRM//McCann, a leading digital advertising agency, where she focused on helping clients understand their customers by leveraging unstructured data with modern NLP techniques. Below, read through some highlights from the hour-long conversation:
Were you able to jump straight into a senior-level position out of academia? What kind of hoops did you have to jump through to land your first job?
At the first job I landed out of academia, my title was "Data Scientist." However, I was the only data scientist in a company of ~200 people, so I felt like I had autonomy and the ability to lead in my role. I did my share of interviewing to get that first job, but in the end, it was worth it. I tried to treat the job search like just another puzzle to solve and get better every time I interviewed or networked.
How did you find the transition going from research into professional work?
For my transition to industry, I distinctly remember that I needed a mental shift more than any new technical skills. The pace of the job necessitated that I didn't always get to spend as much time with certain projects as I would have wanted to. And I was tasked with providing direct, actionable recommendations in how we should adapt our business, which was a bit different than providing results in academia.
When you landed at at MRM//McCann, were you interested specifically in advertising data? And in terms of the team, did you have your eye on a certain fit? For example, did you want an established data team at an established company, or perhaps more autonomy at a newer company?
Prior to working at MRM//McCann, I worked at an advertising agency in Boston, so I was already in the biz. The work MRM is doing in NLP really interested me. As far as finding the right team or looking for autonomy...the answer is YES and YES! I was lucky enough to be on a team of fantastic folks at MRM; meanwhile, I also got to lead my own projects. Both components were quite important to me. I would say that it's always good to ask VERY SPECIFIC questions in the interview depending on what you're looking for in a team and a role.
What was the most difficult part for you in transitioning to data science?
The biggest hurdles for me to overcome were mainly those of changing time scales and my approach to delivering results. The projects I have worked on in industry have been rather fast paced, often on the scale of weeks or maybe a month, which is much faster than the years I got to spend with my doctorate work! I also reframed how I deliver results by making explicit recommendations to stakeholders at my company rather than letting my audience draw their own conclusions. The problems in industry are much more about "how can these results affect the bottom line" and much less about "oh, that's interesting."
What skills carry over from academia to data science?
So many skills carry over! As far as technical skills, many academics have learned about and possibly leveraged techniques from mathematics or statistics. For example, psychology is a field that conducts statistical tests frequently. Many academics also have experience coding, which is a big plus. Academics often have quite a bit of practice communicating technical concepts both verbally and through writing, which is a highly valued skill in data science. And of course the soft skills: it takes quite a lot of "grit" to complete an advanced degree, one of the core attributes we look for at Metis.
What is the most under-appreciated skill for a data scientist to have in your view?
One skill that I think good data scientists have (that some times gets overlooked) is their ability to think logically through a problem. It's not as easy as it sounds! To quickly ramp up in terms of domain knowledge (or at least ask the appropriate questions of someone who is an expert in the vertical) and then apply that subject matter expertise when cleaning data, selecting the model, interpreting the results – it's a complicated process to get right. I think that is one of the most important, but hard to quantify, skills of a data scientist.
HANDLING THE DATA SCIENCE INTERVIEW
What are some of the common questions in a data science interview?
Interview questions these definitely vary from stats to programming to brain teasers. I did see this book recently and have been wanting to check it out.
When you transitioned to data science, especially during the interview process, how did you deal with the case studies and data challenges? Any suggestions for preparing those works?
While the take-home challenges that some companies provide may be time-consuming, I think they are often helpful in terms of learning what kinds of skills the company is looking for – and even helpful for your own education! For example, you might need to use a new type of model or handle a new kind of data you haven't seen before. It's an opportunity to learn! One nice way to prepare might be to ask a friend or mentor to do code review with you. It can be super helpful to have someone else try to read your code and to ofter tips for areas of improvement.
I'm wondering if you could comment generally on how much companies are looking for specific technical skills vs. how employees work and what they can learn. I hear that many companies do indeed look for the latter, but being in a Ph.D. program, it's hard to know whether or not I'm qualified for jobs.
Most companies are looking for some level of technical skills but that varies depending on the company and the role. However, most companies are also looking to hire people that are the right fit in terms of culture and, yes, ability to skill up where needed.
What the average onboarding time for a new data scientist?
Onboarding time can vary, but I will say it is helpful if you can "hit the ground running" and learn as much as you can within the first few months at a new job. The interviews themselves can be quite telling! Every interview is a great opportunity to learn, no matter the outcome.
In your view, do you think it’s necessary to have a data science portfolio to demonstrate to employers that you are capable of doing the job? And if so, how would you recommend building that portfolio?
It definitely helps! Having portfolio projects means that you will have work you can discuss at potential interviews and work that you can point to to demonstrate your technical skills, as well as your tenacity to work through problems and issues that may arise. A portfolio can be built in many, many ways. Coming up with the questions to ask and answer is part of the fun! You could start by taking a look at Kaggle to see the types of problems companies are interested in and then take it from there.
THE METIS BOOTCAMP
I'm curious about post-bootcamp job scenarios of Metis graduates. Being an international student, it's time sensitive for me to land a job after the bootcamp. Normally how long does it take for a candidate to land a job?
As far as post-completion job scenarios, it definitely varies. We have had students land positions just a few weeks after the program ends, and of course, we have also had students take more time and even pass on a few offers before they find the right fit for them.
What are the pros and cons of attending a bootcamp, specifically for academics who've already spent a significant chunk of time and money in grad school and/or postdoc positions?
I think there are many pros! Attending a bootcamp helps 1) skill up in any areas where a student is less experienced (for example, if someone comes from a math background, they may spend time at a bootcamp to improve their programming skills and vice versa); 2) become more acclimated to the quick pace and type of deliverables that will be required in industry; and 3) learn more about the iterative/agile approach that many companies take (starting from a simple model and building it up). Doing a bootcamp does require added investment though (both time and money).
Out of the 5 projects completed in the bootcamp, do you have advice for how to use them to impress employers and increase chances of a job offer?
My best advice as far as selecting a topic for your bootcamp projects is to pick something that really, truly interests you. Pick topics that you enjoy and will *still* enjoy after talking about it many times to interviewers. But, of course, if there is a particular domain that you are interested in exploring, it might be helpful to start working with that kind of data. If for no other reason than to see if you like that field or not!
Learn more about Kimberly and the rest of the Metis team here, and get more information about our Data Science Bootcamp here.