FIU Data Science Bootcamp Application Deadline is Oct. 3 - Apply Now

Metis Approach to Data Science Education (Part 2): Staying Current in a Cutting Edge World

By Paul Burkard • May 30, 2017

Foreword: This is the second entry in an ongoing series detailing the Metis approach to Data Science Education. The series will cover a variety of topics from strategies and philosophy to technologies and techniques, which have been cultivated through Metis’s firsthand experience instructing many aspiring data scientists. The first entry is available here.

This was written by Paul Burkard, Metis Sr. Data Scientist based in San Francisco, featuring an interview with Julia Lintern, Metis Sr. Data Scientist based in New York City.

This past Fall, Metis Senior Data Scientist Julia Lintern spearheaded an effort to revamp aspects of Metis’s Data Science Bootcamp curriculum in terms of both content and content delivery, which marked the first major curriculum overhaul in its 2.5-year history. This blog post is structured as some history/background on the contents of the Metis curriculum alongside an interview with Julia about the changes made.

History of the Metis Data Science Bootcamp Curriculum

Metis began offering Data Science Bootcamps in New York City in the Fall of 2014, and has now run 16 of them in New York, San Francisco, and Chicago. As discussed in the first entry of this series, the project-driven approach has been the rock driving the Metis Data Science Bootcamp curriculum from day one and has remained in place thanks to its success in teaching aspiring data scientists how to actually do the job of data science. The original curriculum content was developed by Datascope Analytics and ranged over general data science topics in Python including data exploration, data visualization, data modeling, supervised learning, regression, classification, unsupervised learning, clustering, SQL and NoSQL, web scraping, APIs, natural language processing, and more. It has been very successful in training hundreds of would-be data scientists over the years and continues to be so.  

However, as one might guess, a lot can change in 2.5 years in an industry as cutting edge as modern data science. As such, while the project-centric style is still front and center, the actual curriculum content has required updating from time to time in order to keep up with the breakneck pace of innovation in the field. Some immediate examples that come to mind include incorporating Big Data technologies, Recommender Systems, and Deep Learning into the curriculum, as well as placing a greater emphasis on Gradient Boosting and Ensemble Techniques. In addition, the Metis team is constantly tweaking and revamping content and content delivery techniques based on firsthand lessons learned. In 2016, it became clear that a “Curriculum 2.0” project was necessary in order to cement the various tweaks that had been implemented over the years. Here is the story of that process:

Interview with Julia Lintern

Paul: How long have you been a Senior Data Scientist/Instructor for Metis, and how many bootcamps have you taught?

Julia: As of July, I will have been at Metis for 2 years. I have co-taught 4 bootcamps so far.

Paul: What was the primary motivating factor for the 2016 curriculum revamp?

: It was the common knowledge between all of the instructors that it was time for an update. The sentiment was that some of the lectures felt dated. Meanwhile, there was so much material we wanted to add. We were really excited about the prospect of adding additional lectures on Deep Learning, Spark, etc.

The primary goal was to ensure a consistent curriculum product bootcamp to bootcamp. In other words, we needed to ensure that each instructor was happy with the curriculum revisions so that he or she wouldn’t have the need to veer from it. In a nutshell, we had to ensure that 12 different instructors concurred with the curriculum updates. (A challenge! More on that later...)

I also wanted to ensure that the new curriculum leveraged and reflected the instructional team's individual strengths and expertise as much as possible.    

Of course, we also had to update the curriculum in terms of content and appearance. Lastly, I wanted to ensure that our revisions were consistent with the feedback from our alumni regarding what they consider to be the most valuable skills on the job.

Paul: Do you think that you met them?

Julia: I think we came as close as possible to meeting these goals. Considering the first goal, it would have been unreasonable to expect that each instructor would have 100% buy-in on every change. However, we were able to reach a consensus on the revisions and that is quite a feat.  

Paul: Are there any topics that have been eliminated or de-emphasized in Curriculum 2.0 as compared to when you first started at Metis? If so, which ones and how/why?

Julia: The developers of the original curriculum did an excellent job and as a result very little of the original curriculum needed to be eliminated. However, the original curriculum was designed such that one entire week was dedicated to D3. Metis bootcamp history has shown that one week for D3 was neither long enough to master this tool nor short enough to allow D3 to be optional for students. We decided to change the time allotted to D3 – in lieu of having one week for D3, we would dedicate one week to data products (including HTML, JavaScript, D3, and flask).

Paul: Are there any topics that have been added or emphasized more heavily in Curriculum 2.0 as compared to when you first started at Metis? If so, which ones and how/why?

Julia: Indeed. One area of content that has been increasing steadily is Deep Learning. Initially, our curriculum included one lecture, whereas now the curriculum contains 4-5 lectures on Deep Learning including Word2Vec, Neural Net & CNN design, and Deep Learning with Keras. There are so many exciting developments in this arena, and of course, we wanted to ensure that our curriculum reflected this.

Paul: What was the decision process like for adding, deleting, or updating content during the revamp process?

Julia: We developed 9 committees to review 9 weeks of instructional content (1 committee per week). Each committee would review their week's curriculum lecture by lecture. The formula that we developed worked out quite nicely. Each committee member (there were usually about 5 of us per committee) would state how they felt about that particular lesson, and we would then vote on whether we would eliminate, keep, or revise the lesson. It was quite surprising how often we were able to meet a consensus - of course, there were extensive discussions and often our meetings ran well over time.

Paul: Besides raw topical content, what other sorts of changes were implemented for Curriculum 2.0? Could you elaborate on those?

Julia: As mentioned, our revisions took the form of adding, deleting, or revising existing lessons. The vast majority of the changes took the form of revising existing lessons. In these cases, we agreed that the topic that we were presenting was relevant, but we wanted to alter the way in which it was presented. For example, we previously taught the subject of regularization via pdf and now it is presented via iPython notebook.  

Paul: Aside from the recent overhaul, what are some other ways that you’ve seen the bootcamp curriculum evolve during your time at Metis?

Julia:  Every instructor had previously added his or her own elements to their individual bootcamp's curriculum over time. One of the challenges (and pleasures) of spearheading the revision was the process of curating this added curriculum to determine which elements should survive to the final stage.

Paul: How do you foresee the curriculum keeping up with cutting edge advancements in Data Science in the future? Do you think it’s important that students are always getting the latest and greatest, or should there always be a certain piece carved out for the core concepts (or both)?

Julia: I believe what makes our curriculum most special is that it strikes the delicate balance of including core concepts while also allowing room and time for teaching state-of-the-art technology from traditional stats to the latest and greatest with Spark. I think this helps us achieve the “largest net” approach, which serves our students best in terms of finding employment.

Paul: What types of advancements do you see for the Metis curriculum in the future?

Julia: I foresee that continual curriculum updates will be required in order to stay current with Big Data and Deep Learning trends. Currently, we have the advantage of providing our students with AWS credits so that they can acquire hands-on practice with deep learning tools and modeling. However, I would like to see the access to a variety of different resources expand to such tools as Databricks and H20.

Paul: Thanks, Julia. Anything else you’d like to add?

Julia: It was great to witness firsthand how the new curriculum fared in real life. I had the chance to co-teach via the new curriculum in last winter’s bootcamp. Although the students were only familiar with the new curriculum and had no basis for comparison, we heard great feedback regarding the structure as well as the depth and range of the content. Further, we were able to see that students gained a strong command of various tools (Flask, Spark, Keras, AWS) faster because we introduced the material earlier and/or because of the influx of material. It was great to see how this gain in knowledge had such strong effects and presence in their project work.


To learn more about Metis Data Science Bootcamps, click here!

Similar Posts

data science
Python Guide: Tutorial For Beginners

By Adam Wearne • July 28, 2021

Welcome to a brief introduction to Python. In this article, we'll provide an overview of the Python language, some of its many use cases, how to install Python on your computer, and how to use Python.

data science
A Virtual Classroom Tour with Course Report

By Carlos Russo • December 16, 2020

During a recent webinar with the Course Report team, Roberto Reif walked the audience through a virtual classroom tour of our new data science & analytics bootcamps.

data science
Course Report Webinar: How is Python Used for Data Science?

By Carlos Russo • September 21, 2020

During a recent webinar with Course Report, Metis Sr. Data Scientist Kimberly Fessel discusses how Python is used for data science, how much Python you should know before starting a data science bootcamp, and more. Watch it here.