Each week for the next 3 months I will be posting a summary of my thoughts and experiences that week in the Galvanize Data Science Immersive. I am currently enrolled as a member of the 13th cohort!
This post will be a short background of how I came to enroll in the program, and why I chose Galvanize over other data science bootcamps.
After completing an MSc in Medical Physics, I was looking to move laterally from research into data science. Although I felt that I had picked up many of the techniques and tools used within data science throughout my MSc, I felt that I was lacking in some of the more applied parts: creating interfaces, presenting work to a lay-audience, telling a compelling story with data that may be interesting to those not directly in your field… Looking to improve on these aspects, I began researching data science bootcamps.
The current players in the data science bootcamp market, in the bay-area at least, seem to be General Assembly, Metis,and Galvanize. I applied to all three, and came to be impressed by different qualities of each program as I progressed through their respective interview processes. Each of the companies had roughly similar interview processes:
General Assembly has a long history of successful bootcamp graduates, but mainly in fields such as web development and user experience; this would be their first data science bootcamp. It was unclear to me whether this being the inaugural data science program should have been weighted as a pro or a con. On the one hand, given that the future success and marketability of the program partially hinges on the success of the first-round graduates, I expect there to be special attention paid to the students. On the other hand, given the lack of prior experience in running the course, I felt there would be high potential for pedagogical hiccups.
Of the three, I felt that the General Assembly interview process gave the best opportunities to showcase softer skills, such as presentation and story-telling. This was most apparent in their take-home assignment in which you are given a SQL database with data related to a music store, and several questions like “how does geographical region impact music taste”, or “what might X store do to improve sales”. After answering the questions, the candidate is to create a PPT, and will present it to General Assembly upon completion. The questions were fairly open-ended, and there was no time-limit for completion - together these allowed for a lot of creativity in the response.
Notably missing from the General Assembly interview process was an interview with an instructor. All interviews were conducted with non-technical staff, so it was difficult to get a handle on the skills of the instructors.
The Metis take home assignment gave 24 hours for completion, which was absolutely enough time to polish the end result. It also gave lots of room for creativity, pushing the candidate to create an idea for a potential project and diagram it end to end (colors and pictures encouraged, words discouraged).
Galvanize had the most technically difficult interview assignment, given 4 hours the candidate is expected to answer a series of questions testing their python skills. The question set was relatively long for that time period, and there were a number of extra credit questions. I opted to complete the extra questions at the expense of cleaning up my code and really polishing the previous results. Galvanize also had by far the most rigorous statistics interview. Overall, I felt there were fewer opportunities for flourish and creativity throughout the Galvanize application than there were in the other two programs.
In the end, I chose Galvanize due to what I perceived to be the strongest reputation online (based largely on student testimonials on Quora), and the rigorousness of the interview. Given the rigor, my hope was that my peers in the classroom would be of the highest calibre.
Galvanize is taught entirely in Python, which I expect will be very interesting for me. I have been using R consistently for the past two years (and have even written a couple of packages for R), and consequently the way I think about data analysis is heavily influenced by several R packages (namely dplyr
, tidyr
, ggplot2
, reshape2
, etc. - the hadleyverse!). Moving to python will force me think in a different way, and I’m looking forward to that!