The Revolution Will Be Live and Personalized
Each year a group of undergraduate Statistics and Applied-Mathematics students compete in DataFest, a live 48 hour data centered hackathon. Most hackathons focus on developing apps, whereas this is more centered on delivering narratives and value driven business objectives which are actionable.
UCLA hosted the event in Los Angeles at their campus, along with a few other local Universities who joined, CalTech being one. Nationally, other major Universities also competed against each other locally. Students were given a few days to explore the data prior to the event. The students were then put into teams teams tasked with delivering some sort of actionable insight for a major company willing to share their data under non-disclosure – this year it was Expedia.com.
It was IRIS.TV’s second year being invited to participate in the event, and it was my personal honor to represent us as one of the roving Data Scientist volunteers and mentors to over 50 teams and 200 local individual competitors.
Over the course of two late nights, I was able to teach some statistical programming ( ~ 95% of the students were using R, versus 5% Python [ go #rstats !! ] ). I was also able to get a feel for this new generation of young talent, share experience and knowledge, and make some new friends in the process.
One of the more sophisticated teams I got to know well, was a sharp group of last-year students who had participated the previous year as well, and were aptly named the “Senior Statisticians.” They were working on pulling in outside data (one of the three primary challenges), and using clustering to define customer boundaries. One of their members just got accepted into the Master of Statistics program at UCLA, another was accepted into their Master of Economics program, and another team member is co-founding a tech startup in the localized discount personalization space.
What I learned most could be summarized on a sticker from a student’s laptop, which he bought from an online sticker store: “THE REVOLUTION WILL NOT BE SUPERVISED (X, Y)” – capturing the essence of this new field of Data Science – a community of hackers and highly qualified math minded people. We are all constantly learning the technical skills and creative skills needed to deliver actionable, interesting, and valuable insights – by whatever means necessary. The quote had dual meaning for me. One, it harks back to a powerful song by Gil Scott Heron, THE REVOLUTION WILL NOT BE TELEVISED. This quote is relevant to a company like IRIS.TV, in my opinion, as we are working hard on the transition to a different paradigm of media, where it will be delivered both Live and Personalized without previous inclination toward a static standard of TV. It also highlights a particular technical point, in regards to Machine Learning – where classification systems employing various statistical programming techniques are generally broken down into either Supervised Learning or Unsupervised Learning – the latter being generally more abstract and bleeding edge, where buckets are defined as a particular subgroup hidden among variables. IRIS.TV employs both Supervised and Unsupervised approaches and values creative insights uncovered from our data via Augmented Intelligence.
Walking away from this inspiring event and the next generation of creative problem solvers, it is clear, the revolution will be live and personalized!
Zecca Lehn is a Data Scientist at IRIS.TV. He’s active in the R&D of our proprietary machine learning applications – e.g., associated with Natural Language Processing (NLP) and asset segmentation, and supports the daily advancement of our proprietary video personalization engine, scaling our infrastructure, and increasing our automated competitive intelligence systems. For nearly a decade, he worked as an independent consultant, quantitative trader, and environmental economist; prior to joining IRIS.TV, he successfully completed NewMet Data Science Bootcamp, and holds all certifications from Johns Hopkins intensive Data Science Specialization.