Why I Quit My Job in Data Science to Enroll in a Bootcamp… for Data ScienceDrew HoppesBlockedUnblockFollowFollowingFeb 6The Road Less TraveledIn November 2018 I quit my job as a Data Scientist at a global research firm to enroll in a Data Science bootcamp at General Assembly full time.
This is an unorthodox path.
Take the below snip from a Data Camp infographic on how to ‘Become a Data Scientist in 8 Easy Steps’ as an example (Easy Steps is an interesting word choice).
Data Camp advice on how to prove you’re a ‘true’ Data ScientistStep 7 gives 2 stars to enrolling in a bootcamp, and 3 stars to getting a job.
The authors seem to imply that there is some ordinal hierarchy in the ways someone pursuing a career in the field can gain experience.
While I appreciate Data Camp listing a few examples of alternative entities where a self-learner can apply new-found skills, I find the placement of the bootcamp within the infographic to be misleading.
A bootcamp would be more appropriately reframed as a choice in the following three educational routes: online program (Coursera as an example), an accredited master’s program, or a coding bootcamp.
Since ‘Data Science’ as an interdisciplinary field hit the scene, there has been a proliferation in these types of programs, and frankly right now the education landscape is the Wild Wild West.
With all of these options that are so varied in their price /curriculum it is difficult to compare the value in one mode of continuing education versus another.
So now that we’ve properly reframed a bootcamp as an educational program, the question remains… why did I choose to enroll in that type of program over another route?.Isn’t a master’s degree better?.Moreover, why didn’t I just stay in my job as a Data Scientist?.Isn’t that the end goal of a bootcamp anyways?To understand my decision making I’ll share my background, where I am now, and what I see for myself going forward.
I hope to share my experience with you, and not to be prescriptive about a path that is best for everyone.
The saturation of articles that talk about the WAY to become a Data Scientist create a tone of authoritativeness, no matter if they contradict 10 other “experts” in the field.
Getting the 1st JobI began the first 3 years of my career working in the private health insurance consulting, working primarily with claims and enrollment data to help employers make decisions about their health plans.
I became comfortable working with data warehouses and loved the details behind how we got our data feeds from the health insurance carriers.
More importantly, I learned how to answer key questions for our clients including: “how would changing from a health savings account to a health reimbursement account affect my bottom line?”, and “if we merged these two companies and chose this benefits plan, how much money would I save?”.
Beyond the questions I could answer, I also loved toying with the presentation of the information.
Towards the end of my time there I started building dashboards in excel that would allow a user to click through a workbook as if it was an application (I wasn’t familiar with business intelligence software at the time and if you’re thinking “… Dashboards in Excel?” check this out).
All of these experiences led me to a field I didn’t know existed until about 5 months before I left my job in November 2015 to accept a position as a Data Scientist.
I was green as pine, but in my interview with them I articulated that I have an analytical mind, a willingness to learn and experience in client facing roles (I was also served in an account management capacity).
It was enough to get the job, but in my mind and I’m assuming theirs, I had a long way to go.
The definition of Data Science was more amorphous 4 years ago (hard to believe, but it was), and companies seemed to be listing every skill known to man for Junior level positions.
Everyone wanted a unicorn and I certainly wasn’t that.
I knew I had to improve at modeling (I was rusty since my Econometrics courses in school) and scripting.
Doing the JobOnce I had the job, the first year was really difficult.
We effectively functioned as a startup within a larger apparatus, and as such nothing really was defined.
In this void during the first few months, I was asked to start training on my own.
Having not programmed since Stata in college, I tried my hand at R.
R had a fairly steep learning curve, especially when nobody in my group programmed in it.
I tried the self-learning route where I would take courses online.
I remember the first course I tried was Roger Peng’s Data Science in R Course through John’s Hopkins.
The first week I felt comfortable and was encouraged that I would be able to make the leap.
After spending years studying and passing insurance certifications, I was used to studying dense, complex material outside of work and balancing it with my everyday life.
The second week of the course was different.
The way that the material compounded quickly and with it being a virtual course I didn’t have a guide to bounce material off of.
While I was struggling to have those concepts click outside of work, inside of work I finally got my first project working on an HIV/AIDS Bureau dataset building a dashboard in Tableau.
There were a couple of developers who knew Tableau in the company, and even just having one or two people made a world of difference for me.
I wasn’t great at first, in fact, if I looked back on my work today it was downright remedial, but it was an incremental improvement in the project’s reporting, and it gave me an opportunity to learn.
By sheer demand for these types of projects throughout the company I gradually eased more and more into a business intelligence role, becoming embedded on projects as diverse as regulatory enforcement for nursing homes and an international health financing project.
I became a specialist in business intelligence/data viz, and largely abandoned the quest to be the big data guru or the machine learning engineer.
I enjoyed the hybrid nature of being in a client facing role, but also getting to dig into the technical aspects of the data on the back end.
While I really got the hang of Tableau and figuring out where it’s limits were, I still struggled with inconsistency in attempting to achieve my goal of learning to program outside of work.
I decided to switch to Python and got as far as completing a few courses in DataCamp, but still did most of my data cleaning in excel, as it was more transparent for others who weren’t familiar with Python.
As time passed, I recognized that on some level the self-learning courses were a challenging fit for me.
I would have streaks where I would be great at programming for 2 or 3 weeks at a time, and then would have a hellish couple of weeks working late, or a string of 3 weddings in a month that would knock me off track.
By the time I came back to it the concepts would be fuzzy and I would have to review from scratch.
Leveling UpAfter scratching online courses as a sole-solution off my list, there were really two options left for me to choose from, a bootcamp, or a master’s program.
When comparing the two, there were a number of similarities and differences.
Neither option had a leg up on the other in terms of having started their programs much earlier than the other.
Data Science as a discipline only really burst on scene 5–10 years ago, so a lot of the programs are still young and changing a lot.
Money was certainly a big factor for me.
A bootcamp price tag ranged from $7K to $16K (not including living expenses), while a data science master’s programs cost as much as a traditional master’s programs and would require me to be out of the workforce for 15 months minimally, or balance going to school at night.
Another factor that was annoying, but a reality was the credibility/prestige in the programs.
Bootcamps embody the spirit of Silicon Valley techno-meritocracy, where anybody with a laptop and a love of learning can hack their way to competency, while master’s programs still had the rigidity and rigor of a structured academic apprenticeship.
In the real world, I had seen that some from academia had a very real hostility to data science as a movement and protested that: “they’ve been data scientist since 19_ _”, “machine learning is just regression under a different name”, and so on and so forth.
There was truth in many of their statements, but also hubris in discarding an entire interdisciplinary field that were bringing together a network of subjects and applications under one linguistic umbrella.
After consideration, I decided to go to the bootcamp route.
From my first few years in the workforce, I learned that it is not where you go to school that has the highest correlation with success, but rather persistence, grit and curiosity.
People with an urge to learn on their own, through meetups, experimenting on their own, reading on their own, were what I wanted to emulate.
After speaking with a few folks who had done the bootcamp route at General Assembly I knew that there was a culture of intense challenges, and people that often leave jobs in completely different fields to throw themselves into the deep end and see if they can swim.
For me to grow, I wanted to be immersed in that culture, with those brave people.
As far as what employers would perceive… if they had bias against bootcamps and were willing to discard an entire talent pool (that takes a full graduate math curriculum) that is certainly their prerogative, and I can do one of two things in response 1.
Try to convince them that I can do anything that someone with a master’s in data science can do, or 2.
Ignore that company.
If hiring from bootcamps are good enough for companies like Deloitte, Amazon, Barclays, IBM, Apple and Boeing (just to name a few), logic dictates that it should be viewed on par as a master’s level program.
Light at the End of the TunnelIt was the best decision I ever made.
In the first two weeks the assignments really tested our ability to use control flow and logic to answer questions with all different data types.
I was amazed how quickly I was able to learn skills that had evaded me for so long.
I had the confidence to write my own scripts and build ETL flows that I could build to automate processes in real world situations.
We worked with Spark and SQL to give us a taste of big data platforms and typical relational databases.
What’s more we were able to apply our newfound programming skills to do inferential statistics, build machine learning pipelines, and build data scrapers to perform classification algorithms (think spam classifier).
Broadly, it gave me the technical skills and the confidence to augment all those experiences from my first position to work with bigger, less structured data and answer questions in automated ways.
For the concepts that we didn’t explore in great depth, the course also gave us the associations using Stack Overflow and google to know the right questions to ask.
As I alluded to earlier in the piece my journey could be viewed as unorthodox.
Frankly I question what “orthodoxy” is in this field.
You talk to a software developer type Data Scientist and they want you to have software development skills, or a big data engineer will want you to be able to spin up Hadoop clusters.
I’m convinced that, at its core, Data Science is like Environmental Science.
You would never hear a Geologist say to a Hydrologist that they aren’t an Environmental Scientist.
There will be all types of different routes Data Scientists will take, and what they will need to know will depend on their domain.
Nobody can be everything to everybody, and my best advice for folks is to take the best route for YOU so that you can provide the maximum value to your eventual stakeholders.
I believe I did.