Choosing a career switch through Galvanize

Cohort #15 of Galvanize SF’s Data Science Immersive program starts May 31! I’ve spent the last year exploring data science on my own and am excited to transition my career from structural engineering to data science. Leaving what I’d consider a comfortable job is pretty daunting, so of course, it took a bit of deliberation to ultimately decide to go with Galvanize. I’ll summarize the process that led me to this decision- hopefully this post (and blog in general) will be a good resource for others considering a career switch into data science.

I'll start off with my background and how I got to data science: My academic background includes a BS (2012) and MS (2014) in Structural Engineering from the University of Illinois at Urbana-Champaign (UIUC). I spent 5 of my 6 years at UIUC as a Research Assistant in the Smart Structures Technology Laboratory, where I primarily explored the use of smart dampers as vibration control devices for large buildings undergoing extreme events (e.g., earthquakes). In the summers, I worked in a variety of other research labs in the US and abroad. Some of the more interesting projects included performing structural simulations of the human eye at a research lab in Texas and working as a National Science Foundation graduate student fellow to develop & implement control algorithms for the National Center for Research on Earthquake Engineering in Taiwan. In a nutshell, I think research is fun.

I left UIUC after my MS and became a Design Engineer at Magnusson Klemencic Associates, a structural engineering firm that prides itself on innovative buildings and “first-ever” damper applications in practice. During my time there, I found myself more drawn to data-oriented tasks, to the point where I began taking Coursera classes outside of work. After completing the Data Science Specialization from Johns Hopkins University and Andrew Ng’s Machine Learning course from Stanford University, I realized that data science was a better fit for my career interests. Everything I enjoyed about my academic research- conducting experiments to find how a system behaves, visualizing & interpreting data, designing & validating an improved system- are key aspects of data science. It’s even more exciting to know that there’s such a wide range of problems I can solve with data, like predicting flight delays or just avoiding masses of hipsters. :)

So I decided to make the switch. It boiled down to three options:

Option A: Continue taking online courses and attending tech meetups outside of work.

  • Pros:
    • most financially stable option
    • most customizable in terms of content and pace
  • Cons:
    • takes longer to develop skills outside of work
    • little-to-no interaction with peers or instructors (difficult to learn by working with others and difficult to build a network)
    • no feedback that what you’re learning is relevant to industry

Option B: Apply to an accelerated program (bootcamp).

  • Pros:
    • learning with peers of diverse backgrounds and similar motivation
    • interaction/connection to companies (good networking opportunities)
    • curriculum based on industry needs
    • shorter career transition time
  • Cons:
    • most programs require a PhD
    • can be pricey for the duration or be extremely competitive for the free/paid ones
    • employers may have mixed feelings hiring bootcamp graduates compared to CS degree holders
    • accelerated pace may result in some content not being learned in sufficient detail

Option C: Go back to school and get a CS degree.

  • Pros:
    • probably easier to secure a job when you're branded with a CS degree from a good school
    • learning with peers
    • being in school long enough to have concepts ingrained in your brain
  • Cons:
    • most expensive option
    • could take a couple of years for an MS degree
    • little feedback on current industry standards or what is relevant to industry
    • isn't specifically tailored for data science
    • programs that ARE specifically tailored for data science (ex. MIDS) are relatively new with limited outcomes information and may only be available online

Option A was out of the question since I had been doing it for the last year already, and clearly it didn't feel like enough. Option C felt safe, but a little slow and boring to me for such a big cost. This post turned me off to MIDS. Option B seemed like the best mix of pros & cons, but I was a bit skeptical of bootcamps. I decided to investigate Option B a bit more.

As a non-PhD holder, Galvanize and Metis were my two most promising programs. I honestly didn't look at Metis much since they were based in New York City, and I hadn't heard much about them as a Seattleite. In contrast, Galvanize hosted all the talks I had been attending and had also acquired Zipfian Academy, making the San Francisco program (Galvanize SF) the longest running data science bootcamp. I applied to Galvanize SF.

Galvanize has a 4-step admissions process:

  1. Submit an online application (includes short answer questions)
  2. Complete a 4-hour technical assessment on Python, SQL, and probability/statistics
  3. Pass a 1-hour Python programming interview in an online collaborative coding environment
  4. Pass a 1-hour technical interview on probability/statistics

The process took ~2 weeks to complete. At the time I applied, I had completed the Data Science Specialization from Johns Hopkins University and was starting Andrew Ng’s Machine Learning course from Stanford University. I took a couple of weeks to learn Python and go through through SQLZOO tutorials before I applied since my coding experience was primarily in MATLAB and R. With this preparation, I was comfortable in the admissions process- the questions made me think and were not overly easy.

I actually enjoyed the interviews and viewed them not only as a way to demonstrate my skills, but also as learning experiences. My interviewers (Galvanize alumni who have continued as Resident Data Scientists) helped show me things like how to write more Pythonic code. They were also happy to answer whatever questions I had about the program, from what they thought the program's best selling point was, to how they handled insurance coverage during the program. Both interviewers said that their top selling points for Galvanize was the network. Since you work intimately with your peers, you learn a lot working with people of different backgrounds. You can also attest to each other's abilities and eventually have contacts at different companies that you actually know, rather than randomly swapping business cards with people at meetups.

After being admitted, I had a week to accept and put in my $2000 deposit (out of $16000... Yeeesh. I've saved money, but I'm still hoping for scholarships). I still had some concerns after looking at the stats Galvanize boasts about on their site.

Does Galvanize employ their own graduates as Resident Data Scientists to inflate their placement rate?

Are the $114k average salary and 94% placement rate facts handpicked from a particular cohort of a particular location?

The Galvanize SF School Performance Fact Sheet provided online gave an overview of 2013 & 2014 program outcomes, which was helpful but a little out-of-date. Apparently, since Galvanize SF is regulated by the California Bureau for Private Postsecondary Education, they have to publish this fact sheet summarizing the last two calendar years by August 1 of each year. This means the more recent program outcomes over 2015 won't be available till August 1, 2016.

I continued looking for outcomes by sifting through LinkedIn profiles, Quora posts, and Course Report reviews. Stalking Galvanize alumni on LinkedIn was useful for seeing what companies people ended up at, but wasn't too helpful for estimating placement rate. It was hard to tell if some people had not found a job yet, or if they just hadn't updated their profile- I just ended up ignoring the profiles with low connection numbers. Quora & Course Report reviews were generally positive, though there were a couple of negative reviews. The positive reviews were more lengthy and in-depth, some even including blogs of their experiences. The negative reviews were shorter, vague, bitter, and generally just didn't have a lot of substance.

Ultimately, what I took from the reviews was you reap what you sow. I concluded that the negative reviews were resulted from students who didn't put a lot into the program (or their review), whereas the positive reviews came from students who worked hard to squeeze out as much as they could from the program. I decided to continue with Galvanize for my career change, and I plan to be one of the positive review writers with a happy success story.

Stay tuned! I'll be including weekly posts about my Galvanize experience- see below for Galvanize's schedule.

Week Topic
1 Exploratory Data Analysis and Software Engineering Best Practices
2 Statistical Inference, Bayesian Methods, A/B Testing, Multi-Armed Bandit
3 Regression, Regularization, Gradient Descent
4 Supervised Machine Learning: Classification, Validation, Ensemble Methods
5 Clustering, Topic Modeling (NMF, LDA), NLP
6 Network Analysis, Matrix Factorization, and Time Series
7 Hadoop, Hive, and MapReduce
8 Data Visualization with D3.js, Data Products, and Fraud Detection Case Study
9-10 Capstone Projects
11 (Break)
12 Onsite Interviews
comments powered by Disqus