💈 Breaking into Data Science — Upskilling & Bootcamps (Ch.2)- 📊

Published in

Ml Ops by Mikiko Bazeley

11 min readFeb 29, 2020

Tl:dr 💁‍️ ➡️ 👩‍💻;

How I went from hair salon girl to data scientist working for an innovative digital health company with no Master’s, PhD, or quantitative college degree. This section covers my attempts at hacking the learning process, choosing a boot camp, and building up my portfolio.

Hi everyone!

So I’m assuming you’re joining me from Chapter 1 of my story which talks about my background and early professional career before committing to becoming a data scientist. If not, please check out Chapter 1: Pre-Data Science where I talk about my motivation and early professional experiences (including “Funemployment”, hair salon desk girl, and connoisseur of early-tech start-ups) before determining data science was my dream career.

Becoming a data scientist isn’t about having the right degree or education.

It’s not about how many MOOC’s you put on your linkedin or how may repos you have in your Github (although I will admit, listing certifications and projects on your linkedin profile is an easier way to keep track of them other than a google doc).

A successful data scientist is ultimately someone who solves problems.

The problems might be of a particular nature and structure, requiring the use of mathematical formulations, programmatic knowledge, and domain expertise.

But that doesn’t mean that a specific degree, graduate program, certificate or specific project will equip you with the problem solving capabilities that are really important.

Chapter 2: The Uphill Climb to Upskill

The Big “No”

The mind that opens to a new idea never returns to its original size. — Albert Einstein
Research is formalized curiosity. It is poking and prying with a purpose. — Zora Neale Hurston

After about 6 months I had the chance to interview for an internal data scientist role on a different team and completely bombed the interview (I know, so brave for admitting failure right?). During the post-interview debrief I asked one of the senior data scientists what I could have done better. “Could I have displayed a better attitude?” “Could I have studied more on specific topics?”

After silently looking at me for a few minutes, she said:

“With the state of the data and data infrastructure such a mess here, you really need to be at a senior level to do well … at times it’s hard even for me to manage the different work streams across shifting priorities, teams and projects”.

She encouraged me to try to find a place where I could work on data science-y projects end-to-end, get more hands-on experience and have the time to also attend a boot camp or academic program on the side. I absolutely needed some kind of formal learning structure and environment as hacking the process wasn’t working out.

A couple months after our discussion I made the strategic decision to transition to a company that had a single central data warehouse, a significantly smaller product portfolio and smaller data team.

I took a title cut and moved to a role where I’d be able to build out my portfolio and have total autonomy over my projects (including end-to-end model building, from project scope to feature engineering to model evaluation and deployment).

There were definitely people in my network that couldn’t understand why I would leave an established company (with a great benefits package & salary) for a lesser known early stage start-up.

Even my family, friends and closest co-workers asked me what I was thinking (correction, whether I was thinking).

“Isn’t a start-up risky?”
“How do you know you picked the right one?”
“What if you regret the move?”
“You just got through the hard part of the first year, why would you just go and leave your budding network?”

Thankfully though, I had experience working for start-ups earlier in my career, which helped me understand that risk is relative and that I could survive the boom-bust cycles of start-up land.

Given the informed risk I was taking, the answer to me was pretty clear: I needed to keep pushing forward. I needed to get my hands dirty with valuable projects, working with data and building models. Most importantly I needed to make a conscious decision to invest very deeply into upskilling in a structured environment and commit to the journey.

Evaluating my options & Choosing a boot camp

A system of education is not one thing, nor does it have a single definite object, nor is it a mere matter of schools. Education is that whole system of human training within and without the school house walls, which molds and develops men. — W. E. B. Du Bois

In putting together my criteria for a data science program I needed to consider the following:

Length of program: Keeping momentum up would be key to continual progress, so preferably =< 6 months.
Academic (Master’s) vs Bootcamp: This was an easy one as basically all the grad schools I called said “no” (apparently no one’s in a rush to enroll a dunce cap wearing, 2.3GPA wielding troglodyte). So boot camp it was.
Full-time or Part-Time: The opportunity cost of a full-time program, aside from the cost of the program, included missing out on earned income + gained experience in that time. (San Francisco rent is also insane)
Cost & Payment Plan: Cost needed to be less than $10k. Springboard came in at $1.5K per month with the monthly plan which was incredibly affordable, especially on a cash basis.
In-person or Remote: Although I really enjoy in-person classes traffic & commuting in SF is nightmarish. Remote also meant I could structure my study around my work hours.
Program Structure: Essentially I identified 3 key components as being important to me: mentorship, project development, career assistance. A program needed to meet these (but not much else).

To summarize: At less than $1.5K per month (spread out over 6 months), with a weekly mentor check-in, and decent course material (curated from external sources), Springboard seemed like a great option. I would also have two machine learning projects that I could take to potential recruiters. An additional benefit was that as part of the program Springboard partners with DataCamp to give you access for 6 months after the program ends.

Overall Springboard proved to be a great investment and has entirely changed my outlook on the value of non-academic boot camps or certification programs.

Springboard’s program focused on python and I went from having never, ever, ever touched python to learning how to leverage core data science libraries like numpy/pandas/sklearn/seaborn/seaborn, being able to roughly describe various tasks & concepts in machine learning and build some decent classical machine learning models.

My Personal Apprenticeship

A map does not just chart, it unlocks and formulates meaning; it forms bridges between here and there, between disparate ideas that we did not know were previously connected. — Reif Larsen

As part of my personal apprenticeship I focused as much as possible on using the tools and skills I learned in the curriculum to initiate and execute data science projects at my company for real-time hands-on apprenticeship-like experience.

For example, for my ‘Sales Classification’ model, I needed to access our redshift instance using SQL Alchemy, meet with our data engineer to track down data and confirm sources, talk to our Salesforce team to understand if additional data could be pulled upstream from our SFDC instance and talk to sales ops to understand potential opportunities for further data enrichment through external data apps. Once I completed a first pass of the model I also needed to socialize the findings with my core business partner team, present on the model performance and methodology to the BI team, and utilize model interpretability (in the form of partial dependence plots) in order to speak through the feature importance and impact on predicted labels with VP of Marketing.

I would work on the model during lunch or after everyone had left the office, review the progress with Rajiv Shah at the end of the week, read through various tutorials and sites to solve problems as they came up, and then update my business partners on any interesting findings.

My mentor Rajiv Shah would give me valuable feedback on my capstone projects as well as on how to interface with key stakeholders, point me in the right direction when I needed additional learning resources, and would cheer me up when the impostor syndrome would rear its ugly head up.

How my week typically looked. My weeks were booked weeks in advance, no room for anything or anyone that wasn’t work, working out, study, family or date night time.

By the numbers:

Hours spent on the boot camp (studying/working on assignments/projects) per week: 5~15 hours (could be as much as 20 hours when debugging and refactoring).
Duration of bootcamp: 8 months (initially slated for six months).
Artifacts produced: Two capstones and at least six mini-projects.
Other activities during this time: Working full-time (M-F, 40 hrs per week), training at gym (4–5 days X week at 3 hrs per session), weekly family dinner (Saturday nights), date nights with boyfriend (two nights per week + one brunch). Occasional movie nights with my boyfriend and/or friends or work trips. There wasn't room for anyone or anything else during this time including close friends.
Study times: Study would occur around lunch time (12p-1:30p), end of day (4p-5:30p) with some more study occurring at night after the gym between (9p-11p). Work happened around those times.
Materials Consumed: As part of the Springboard curriculum I also completed about 22 DataCamp courses, watched at least 30 hours of videos (probably a lot more), and read at least 100 pages of articles directly recommended or necessary for solving the assignments.

Final Thoughts on my Boot Camp Experience

I thank whatever gods may be
For my unconquerable soul.
…
I am the master of my fate:
I am the captain of my soul.
— “Invictus”

Although the program was estimated to take six months, I needed to take another two months to complete all the projects.

However I’m proud of myself for taking on the challenge, doing the best I could, and always trying to push the envelope with my learning.

Prior to Springboard I struggled to complete MOOC’s and online courses or programs.

Not only did I finish but while I was working on the boot camp nights and weekends, I was also spending meaningful time with my family and boyfriend, taking care of my body, working full-time, and learning to cook (along with doing extra readings and study related to data science).

At any point in time I could have given up and instead I dug deep, rearranged my schedule, stayed up late, work up early and disciplined myself to make time to complete the assignments.

Areas that I particularly grew in as a result of the program:

Applied statistics & probability theory: Traditional statistics had been my weakest area so what I needed to do was cover half my bedroom with whiteboard wall paper, take my notes and rework them until I understood the logical structure in the context of solving product and marketing problems.
Learning to stay focused: I had a really bad habit of rat holing when it came to feature engineering or EDA. At some point you have to draw a line in the sand and say “enough is enough”. You’re there to solve problems using data science and machine learning, not to tinker.
Understanding NLP: My biggest struggle was understanding the different algorithms and tasks in NLP. I have no background in linguistics, programming or symbolic languages so I needed to read and re-read every sort of explanation and code example out there.

Mikiko Bazeley on LinkedIn: #DataScience #WalkMe #transportationanalytics

So happy to say that I finished Springboard's #DataScience track! It was a challenging & highly valuable experience…

www.linkedin.com

If you’re interested, you can find out more about my capstone projects here:

Classifying Sales Calls: Github Repo
Classifying Kickstarter Campaigns Utilizing NLP Feature Engineering Techniques: Github Repo

I’ve also documented my Springboard experience and coursework in a single repo as well:

Springboard Data Science Career Track: Github Repo

Resources that I used to upskill, aside from Springboard’s Data Science Career Track, included:

DataCamp: They offer a number of tracks (which are curated collections of their courses)
Kaggle Learn: Kaggle now offers some notebook based courses on everything from Feature Engineering to Model Interpretability
PyCon Youtube Channels: PyCon 2019 and PyCon 2018

Additional resources are documented in the wiki for my Springboard repo (which is a WIP).

If you’re interested in checking out Springboard (and getting $500 off your tuition), feel free to click through to my referral code.

By the end of the program I had grown my skill set, built out a portfolio with two projects, gained confidence and utilized what I learned to implement projects at work.

The next big challenge was to throw my hat in the ring and navigate the job market.

And if I thought my previous job hunts were tough, I was about to learn how brutal the data science job hunt could be.

Thanks for reading this far!

If you’re interested in learning more about my experience with the data science job search process, I write about my experience applying and interviewing for jobs in Chapter 3.

Interested in following along with my Data Science journey? One of the easiest ways is by connecting through LinkedIn. Feel free to send me an invite and let me know what you thought about my series! And if you really love what I wrote, consider buying me a coffee at www.buymeacoffee.com/mmbazel & keep me writing! ☕