Published on July 31st, 2014 | by Travis Korte0
5 Q’s for Ryan Orban, Co-Founder of Zipfian Data Science Academy
The Center for Data Innovation spoke with Ryan Orban, Co-Founder and CEO of Zipfian Academy, a Bay Area-based data science school. Orban discussed using software development techniques in an educational setting and how Zipfian determined its curriculum of essential data science skills.
Travis Korte: A lot of colleges and universities have started offering masters degrees in data science and analytics. How do Zipfian Academy’s programs differ from these offerings? What kinds of students might not want to go for a full degree?
Ryan Orban: Zipfian Academy’s data science training programs focus on hands-on problem solving with practical, real-world applications. The skills we teach and the technologies our students use are based on input from industry and reflect the cutting-edge of techniques used in the field. That’s very different from traditional colleges or universities, which typically focus more on theory and less on practice.
Academic institutions take pride in publishing research, and job placement services are often viewed as a nice-to-have. At Zipfian Academy, they’re critical. Many of our graduates receive multiple job offers from employers within days of completing our program. Our deep connections with companies in our hiring partner network—Facebook, Linkedin, Tesla Motors—would be extremely difficult to replicate in academia.
We’ve built highly specialized curriculum that integrates subjects across different disciplines in a way that reflects how data science is done in the real world. Many of the new data science programs popping up around the country are simply repackaging classes from other departments—computer science, statistics, machine learning—and not providing adequate context for how they fit together. Data science is more than the sum of its parts. We’ve seen first-hand that it requires new paradigms of instruction that are not typically not available in traditional academic settings.
The program is structured such that students spend their days implementing real solutions to common problems faced in industry. We use innovative teaching techniques like pair programming, collaborative code review, and test-driven assessments to maximize understanding. Lectures are kept short to maximize hands-on learning and personalized instruction.
TK: What do you look for in an instructor?
RO: Instructors at Zipfian Academy are data scientists and passionate educators who apply the same creative approach to teaching that they apply to data science problems. We look for many of the same qualities we look for in applicants: prior programming experience, statistics and math background, excellent communication skills, and a curious yet pragmatic approach to solving problems.
TK: How did you determine the curriculum? How did you decide what the most essential skills for a data Scientist were?
RO: From the start, we knew the key to building a compelling curriculum was focusing on industry. We collected input from some of the most renowned companies in the space: Facebook, Linkedin, Eventbrite, and other data-driven startups. By interviewing data scientists at these companies, we discovered that data science means very different things to different people. Each company is using data science in a unique way to increase revenue, improve products, and generate insights about themselves and their customers. The trick was to find a common thread that is essential for all data scientists to have, creating a solid foundation on which to build the specific skillsets required by each individual company. We arrived at our current curriculum through an iterative process of improvement, taking into account input from our hiring partners, students, and industry advisors.
TK: Where do Zipfian Academy’s students end up? Is it just the Bay Area Tech Industry, or do students use Zipfian as a stepping stone to other academic programs as well?
RO: From the first two cohorts at Zipfian Academy, we’ve placed over 91 percent of our graduates in data science roles, with titles like Data Scientist, Data Engineer, and Machine Learning Engineer. Examples of companies where our alumni work are Facebook, Tesla Motors, and Change.org.
There are always data-driven startups in the Bay Area interested in our talent, since they employ more individuals in these roles than any other companies. We’ve also seen interest from New York, London, and around the world as the demand for data science talent continues to rise.
We see Zipfian Academy as an outcome-based program. We measure ourselves by the quality of our graduates and their outcomes, be it a full-time job, internship, or returning to their company with an enhanced skillset. Most of our students are late in their career, either as academics or professionals, and are looking to make the jump into a full-time data science or data engineering role.
Interestingly, regarding whether Zipfian Academy is a stepping stone to other academic programs, we’ve actually seen the reverse. We see many applicants who have recently finished other programs, such as the Data Science Specialization on Coursera, the Data Science Course at General Assembly, or a web development intensive like App Academy or Hackbright. This preparation isn’t necessary for all students, but it can help tip the scales during the application process.
TK: If your current offerings are successful, do you foresee offering more specialized (say, industry- or sector-specific) tracks in the future?
RO: We just launched our Data Fellowship and Data Engineering Immersive programs this spring. The Data Fellowship is a 6-week program for quantitative PhDs and data researchers looking for help transitioning into industry. It’s free of charge and focuses on filling in knowledge gaps, building a portfolio project, and gaining insights on how they can apply their skills in an industry context. The Data Engineering Immersive launches in January 2015, and answers the growing need for individuals who have experience building distributed systems to transform and extract insights from large datasets.
There’s a much more in-depth answer to this question here.