After spending years in academic research, Justin Knight wanted the skillset to share interesting insights from data. He dabbled in experimental data-driven artwork, then officially transitioned into the data science industry by attending Springboard’s online data science bootcamp. After honing data science skills like SQL, Spark, and D3.js at Springboard, Justin tells us how his final project helped him land a job as the Principal Data Scientist at Nielsen, where he helps improve business practices for Coca-Cola!
What's your pre-Springboard story?
I have a bachelor's degree in psychology and a Ph.D. in cognitive psychology with an emphasis in cognitive neuroscience from the University of Georgia. I studied EEG (Electroencephalography), which measures brain waves and how it relates to different cognitive processes like memory. I really enjoyed the research, and continued studying EEG in my Postdoc at the University of California, Davis, along with functional magnetic resonance imaging research (fMRI), also relating to human memory.
In my research, I worked with large amounts of data and enjoyed data analysis. After my Postdoc, I went on to do a research assistant professorship at the University of Georgia where I worked with people with schizophrenia and bipolar disorder to understand how their brain rhythms are different from people without those disorders.
Why did you decide to transition out of academia and into Data Science?
I enjoyed all my research, but I wanted to see my research efforts and results having an actionable impact. I was successful at publishing papers and that's where it stopped. Unfortunately, everyday people aren't reading scientific journals.It was a little anticlimactic and made me yearn for something else.
I ended my professorship and explored a hobby doing science-inspired arts. Directly prior to Springboard, I was making digital artwork where I layered drawings, photographs, and mathematically accurate plots, created with simulated data. I wanted to show scientifically meaningful principles through interesting and aesthetically pleasing art. I realized it was more of a hobby and I missed doing hardcore data analysis. That's what drew me back into data science.
I specifically wanted to attend a data science bootcamp to get practical industry skills. I had heard that academia prepares you about 90% for industry science jobs, but you’ll need those additional 10% skills like SQL, Apache Spark, etc. That led me to Springboard.
Did you research other coding bootcamps? What stood out about Springboard?
I found it through searching online for different approaches for transitioning from academia to industry. I did do some research on other bootcamps on Course Report, which I definitely found helpful. I looked into The Data Incubator in New York and Insight Data Science – those are geared towards people that have been in academia and have Ph.D.'s. I also looked at General Assembly.
When making my decision, it was a combination of cost and timing. I was able to start Springboard within a month, whereas Insight and The Data Incubator were several months down the road. Beyond that, the others were way more expensive at around $14,000 for 8-40 weeks, whereas Springboard offered a self-guided pace so you could choose to pay as you go. I finished Springboard faster than the expected six months.
I was able to pay less and also work remotely from Athens, Georgia which was a key part of my choice. The other bootcamps would involve me moving which wasn't feasible for me at the time.
Springboard has a job guarantee. Did that contribute to your decision at all?
Actually, yeah, that did have an impact. I researched the other bootcamps through Course Report and on their own sites, and most of them had pretty high placement rates for students. Having that guarantee was nice.
Was it hard to get into Springboard? What was the application and interview process like?
The application process was fairly challenging. They said about 20% of applicants are accepted to Springboard. I did a basic probability and statistics test to show that I had that basic knowledge. I think Springboard has some folks do a coding challenge as well, depending on your background. Since I already had Python and MATLAB experience, which I used mostly in academia, Springboard felt comfortable with my programming background. There was a video chat with one of the admissions consultants in addition to the test.
How many people were in your online cohort? Did you interact with them at all?
The interactions were more like posts to message boards online. I would imagine there were around 10-20 people in my cohort based on my interactions. Springboard starts a new cohort at the end of each month, so we were able to connect on LinkedIn and build our network.
There were weekly class meetings that you could attend with other cohort members, but because I was based on the East coast, it didn’t align with my schedule. However, since the course was self-paced, I had a regular weekly mentor meeting, which was a great component.
How long did it take you to complete Springboard since it's self-paced? How many hours did you work per week?
It took me four months to complete the course and I actually had some different stoppages along the way. The course is definitely feasible to complete in three months if you're able to devote all of your time to it – and most of the time I was working on it full-time. My goal was to transition into a data science career as soon as possible.
I'd say it was a typical work schedule – 30 to 40 hours a week on average.
How did you stay engaged and motivated while learning online?
That's a very good question and there's definitely a challenge that you face in this type of online environment. I made sure to keep focused on my goal: transitioning into a full-time job as a data scientist. I also picked projects that were of interest to me. Topics that I cared about kept me coming back – even though it did get tedious and technical at times.
In any kind of task like this, there are times where you face shortcomings, or the analysis doesn't work out. That’s a challenge that happens in industry and academia alike. You have to be resilient and know it's just a minor setback – you still learned something. Even when you get an error, you learn what doesn't work so you can then try the next approach.
Tells us about a typical day at Springboard. When you logged on – what happened?
Since Springboard is totally self-paced, you can see the entire curriculum from Day One. You can choose which parts of the curriculum to work on any given day, which is nice. It’s another way to avoid getting burned out. Let's say you chose to do a number of different coding projects and want to take a break from projects – the next day there are videos that you can choose to watch.
Some of the Machine Learning courses came from Harvard's online data science courses. There were also different articles to read about updating your LinkedIn profile and finding a job. They try to intersperse technical videos, coding, and career prep throughout the course. I generally would follow the curriculum, though at times I would definitely jump ahead or pick certain videos to keep my knowledge fresh.
How many instructors or mentors did you work with at Springboard?
I met with probably four to five mentors on a fairly regular basis. We were aware of the other mentors that you don't regularly talk to but have access to.
Tell me about your capstone project at Springboard.
My first capstone project was my favorite and I put in a lot of time into it. You're asked to produce three different capstone project ideas. When you're choosing your project, you write a one-page outline of the data set approach that you will use and ideas for potential clients.
I chose to build an an NFL play-by-play prediction model that predicted the outcome of the next play to help coaches and defensive coordinators make data-informed decisions about what player should be on the field, what play they should call, and how they should line up players. It can also be used in fantasy football where daily fantasy players could have a better idea of which teams, depending on their opponents, will be more heavy in runs or passes in the coming week.
What tools did you use to build your capstone project?
I accessed data from a nice online database that actually pulled from the NFL website and I organized it into a PostgreSQL database across eight different tables. I did multiple pulls to aggregate and merge that data into a Pandas dataframe in Python.
I focused on pass versus run and trained a number of different base algorithms like random forests, support vector machines, neural networks, and gradient boosting, but performance was leveling out around 69% to 70% accuracy. I was able to boost performance another 4% by ensembling eight different base models that were diverse in their predictions. It was something that I was quite proud of. In all of my job interviews, I got amazing feedback on that project.
You came to Springboard with a lot of technical expertise. What were the new technologies that you learned?
I was new to working with real-world data and showing off my skills – I learned how approaches in academia translated to real-world problems.
Learning SQL programming and accessing different SQL databases was something that I gained new expertise in. Also, I used tools like Apache Spark to do more distributed big data processing in memory across different computers. Doing some more interactive data visualizations with Bokeh and D3.js were other things that I hadn't done before.
Did Springboard help you find your new job?
Springboard helped with career prep and the job search where we were exploring different companies I'd be interested in. We also did mock interviews, technical interviews, and take-home coding exercises. They put me in contact with a few different employers. They didn't specifically connect me with my new job at Nielsen, but the training definitely helped. I connected with Nielsen over LinkedIn and had several interviews which were ultimately successful.
What is your role at Nielsen? Is the job what you expected so far?
I started about three weeks ago and I'm still getting access to things, but it's definitely been a positive experience. I feel better prepared and confident in this role given my training and experience through Springboard.
I am the Principal Data Scientist at Nielsen in Atlanta. I work about four to five days a week onsite at Coca-Cola where I am building different statistical and machine learning models to understand Coke's sales data and some of their customer surveys to better understand the predictive variables impacting sales and shipments.
My other role at Nielsen involves merging data from their different surveys and sales data that they get from retailers which is more in-depth than Coke’s data. Coke is tracking their products whereas Nielsen is tracking everything that goes out of supermarkets, stores, Walmarts, and things like that across the country. So we're looking to merge that data and build better-informed predictive models of consumer purchase behavior.
Did Springboard help you communicate business insights to clients?
That was definitely touched upon. Besides the main projects that I worked on, there were also about 15 different projects. Once built, you had to give recommendations to the client. So that was definitely a skill that was encouraged.
Also, my past experience in academia of writing papers, then selling my research to journals and to reviewers, definitely helped. Science is about telling stories from data so it’s helpful to know how to take your analysis and programming to a level where it's understandable to others.
How was your first month in your role at Nielsen?
I jumped right in. I work on a three-person team that has a bit more business experience than I do – it’s nice to pair with my technical experience in analytics.
We're paired with a 4-person data science team at Coke. I've had regular meetings to get an understanding of what kind of analysis and modeling they've done so far and brainstorm new approaches to build more robust models to gain a better understanding of the data. Staying in frequent contact, updating one another on the progress of projects, making different project plan metrics, assessing performance, and making sure we're on the right track.
Looking back on the last few years and your career change from academia to industry, do you think that you would’ve been able to make this change on your own without Springboard?
That is a good question. It is possible, but I'm convinced that it wouldn't be an easy transition if I were still in my Postdoc. That period where I worked as a freelance artist also added to my need to do a data science bootcamp.
I know a number of former Postdocs who did data science bootcamps straight out of their Postdoc. Some good friends of mine are actually considering making the jump and doing a data science bootcamp as well. Even though you're technically proficient and learning a lot of the same skills, there's still that concern that you haven't been in the industry. Employers want to see you work on certain types of problems with certain tools to be more comfortable in hiring. I definitely got way more responses from employers after my experience at Springboard.
What advice do you have for others who are thinking about making a career change and attending a data science bootcamp?
One – be clear on what your goals are and have that picture in your mind because there are going to be rejections along the way. I certainly didn't get an interview from all the job applications that I sent. You have to be resilient, persistent, and push through any setbacks that you may have. Even in your data science bootcamp work and the projects you do, you need to keep your head down and push through. Pick ideas, topics, and projects that are of interest to you to keep yourself engaged.
Two - keep in mind that it's a long process. Consider DJ Patel, the first Chief Data Scientist of the United States – it took him six months to get his first data scientist job.I hear many stories about people transitioning from academia to industry, and they expect that it will take six to nine-months to land a job. I fell right into that six-month mark myself. Keep your head down – it's a long road but if you put in the work, it'll definitely pay off.