One of the most frequent questions we hear, right behind “so, what exactly is a data scientist” or “what makes a great data scientist”, is “how do I become one? I should probably just get a Master’s, right?” Perhaps not anymore; rising costs, changing demand, and the Internet are disrupting this traditional path and providing two viable alternatives. At the other extreme, self-learning through Massive Open Online Courses (MOOCs) give access to courses at an extremely low cost (often free), but leave it “as an exercise for the reader” to identify a suitable set of courses and tools to round out a coherent skill set. In the middle ground, Bootcamps offer students a structured learning environment at a far more affordable rate compared with obtaining a Master’s Degree. So, “which path do I take?”
The triple life of Audrey “I want to be a data scientist” Heepbourne
Let’s do a thought experiment to compare the three experiences. Let’s say Audrey Heepbourne has a four-year engineering degree from State University of New York. Audrey’s been working for a few years but is not particularly excited about her job. She’s excited by quantitative analysis, occasionally writes macros in Excel spreadsheets, and heard that “data science” is sexy but doesn’t know how to get started or even if she has what it takes to be a data scientist. (We see this in many potential data scientists, all she needs is a nudge but she thinks she’s worlds away. Believe in yourself, Audrey; be bold!)
To help illustrate the different paths Audrey can take to train in data science, we describe in detail below three hypothetical scenarios. If you’re running out the door to catch a bus that’s leaving in 5 minutes, here’s a table that summarizes our assessment:
Table 1: Summary of different options for training in data science
In the Master’s scenario, Audrey takes a lot of core classes and some electives over 1-2 years (~$60,000), all of which teach her a lot of the theoretical knowledge related to data related algorithms. The courses have lots of problem sets and homework assignments that she works on with her classmates. A few classes have final projects that are designed to assess whether Audrey understands the theoretical knowledge but do not simulate her responsibilities at her future job. She enjoys socializing with the 20-30 other students in her program while participating in other aspects of campus life, gaining one or two life-long friends along the way.
Her professors are exceptionally gifted scientists that know the ins and outs of data analysis quite well, but most of them have never used data to solve applied problems in industry. A lot of them are good teachers with experience and expertise on the theoretical side of data science including proofs behind machine learning algorithms, analytic derivations of formulas for statistical tests, optimization of different data structures, etc., rather than industry experience of designing data science solutions for companies.
After completing her Master’s degree 18 months later, Audrey can add her 3.7 GPA, a project or two, and maybe even a summer internship to her resume. She attends the university-wide campus recruiting day and, interviews with a couple of companies, and ends up joining the data science team at the company where she interned during the summer. She spends the first six months at her new job learning how to connect a lot of the theory she learned with the day to day challenges of her projects at work. Audrey's solid academic background continues to help her a lot with upcoming projects beyond that as well. However, she will still hit problems where she doesn't have the necessary knowledge yet. She doesn't know everything (and nor does anyone else).
In the self-taught scenario, Audrey spends a considerable amount of time just googling around and learning the skills that she needs to have in order to be a data scientist. She has to learn about the pros and cons of R vs python vs Excel vs SAS, what the word “hadoop” means, what databases are and whether they are relevant for her needs. She also explores the world of MOOCs that are organized by Johns Hopkins, Stanford, and beyond. After taking two MOOCs at a time for the first cycle of classes (4-5 hours per week per class), she realizes its a bit tricky to juggle her courses and work and decides to reduce her course load to one MOOC at a time. Perhaps she does this during her current job or perhaps she heads to the local café after work, but in either case it is largely a solo effort over 1-2 years.
She finds the video-recorded MOOCs really interesting—Andrew Ng and Roger Peng are total badasses—and that its really convenient to be able to skip the parts she already knows. She’s learning a lot but wishes she could ask questions live during lecture rather than having to pause the video to google what the word “responsibilities” means in the context of a “gaussian mixture model”. Beyond the MOOCs, she has found a really great python tutorial and is starting to build up enough programming experience to experiment with scipy, pandas, and scikit-learn. As her confidence grows, she signs up for a Kaggle challenge and optimistically thinks she can place in the top 5 after working on it for a weekend only to realize that it takes a lot more effort than she had previously thought to optimize her solution to be a top contender. In her spare time, she searches around and reads up on how her learnings are applied in practice, and starts experimenting with what she’s learned to analyze her retirement portfolio and to mine the tweets of all the people she follows.
After taking the 9 certificate courses from Johns Hopkins and shoring up her Machine Learning with a few other classes, she earns her data science certificate and adds it to her LinkedIn profile. After four months of searching for a job, she is contacted by a recruiter who ends up having a job that’s a perfect fit for her not too far from where she lives. She continues to learn about python, data visualization and machine learning on the job with her team and starts to adapt her skills to solve practical problems. When she hits a road block, she is totally confident that she can find the answer because her Google skillz are beyond reproach.
In the bootcamp scenario, Audrey enters a 3-month bootcamp (at ~$10,000) that is centered around completing one or more projects that are designed to replicate common projects in industry. While the focus of the bootcamp is on team-based project-based work, she also learns the appropriate theory behind her analysis and why it is relevant to this project. By getting a guided tour of relevant theory through practical application, she is applying her theoretical learnings right away. She enjoys the company of her teammates and the thrill of intensely focusing in a collaborative way with her peers, some of which she suspects she’ll stay in touch with beyond the bootcamp.
Since Audrey’s instructors at the bootcamp come from industry, the focus is constantly on designing solutions that work in practice. Over and over again, she faces practical questions, like “My project just hit a dead end. For predicting if a given patient has the disease or not, I tried the naive Bayes algorithm, and it’s doing terribly. I’m not sure why. What should I do now?” Her instructors guide her towards other classification methods that would be more suitable to solve this type of problem. Audrey is able to learn about new tools and immediately put them to practice but this inevitably leads to further questions. The process is cyclic and happens repeatedly. Through the help of her instructors, Audrey is becoming and expert in learning and applying relevant theory quickly while making efficient tangible progress on her project.
After graduating, she has proof of her skills in the form of a portfolio she can show to potential employers to prove her value as a data scientist. Taking advantage of the Career Day and the Talent Placement Manager that the bootcamp provides, Audrey lands a coveted position within a month after graduation. Her first month in her new job feels like a continuation of the bootcamp itself. Audrey doesn’t know everything but she is equipped with the skills to quickly learn, adapt, and progress with her work project.
A love letter to bootcamps
As technology increases the rate of change of society, the most successful workers will be those that can quickly shift to new specialties and learn on the job to meet market demands. The bootcamp provides the benefits of personalization, credentialing, and social learning that a Master’s degree offers, but at an accelerated rate with experiential learning. Sure it is more expensive than being self-taught, but the connection with employers and the guided, experiential learning process increases your confidence to tackle the uncertain prospect of making a career switch. To become a data scientist, you don’t need to have postgraduate degrees, or 20 years of experience, or be proficient with every data-related technique and tool under the sun. What you need is to have enough baseline knowledge and experience, and the skill to constantly adapt and learn. Bootcamps, in our opinion, are the perfect medium for making the transition.
Full disclosure: we designed a bootcamp
Since we love the bootcamp format so much, when Metis (a part of Kaplan) contacted us about partnering to design a data science bootcamp, we jumped at the opportunity. We thought we could take all these points we see as the advantages of the format, and elevate them as much as we could. So, we designed a course that would give Audrey a lot of experience with 4-5 projects, and a guided route of several core data science concepts and approaches. She can quickly build the necessary foundation without the burden of teaching herself everything or paying the handsome price of a Master’s program before realizing her dream job.
Metis data science bootcamps are available in Chicago, New York City, San Francisco, and Seattle. Learn more about it here if you’re interested.