CV:
jao holds a Ph.D. in theoretical Physics (Barcelona) and a BS in Computer Science (Madrid), and he turned to computers in the 90s working for telecoms and aerospace (Indra), AI and web security startups (Isoco, Scytl), our internet overlords (Google), next-generation user interfaces (Oblong), with an interlude in academia teaching programming and networks (Universitat Autònoma). jao co-founded BigML in 2011, where he's CTO and hacker-at-large. He has been involved since the 90s in Free Software projects and programming languages (GNU MDK, Geiser, xmobar, etc.) He loves Scheme, books and he is learning to play piano.

Abstract:
This talk is about our experience in BigML, a startup creating a Machine Learning platform geared towards intelligent, distributed processing of big data in the cloud, offered as a SaaS platform for the non-initiated. We'll see our current platform in action and discuss the algorithms and machine learning techniques (such as decision trees) that are used under the rug, as well as the knobs and fine tuning available to advanced users of the service.  We'll also touch upon our technology choices (such as clustering, messaging or storage systems) and non-choices (e.g. Hadoop) and how they've been influenced not only by technical constraints, but also by the need to quickly evolve and transform the system as we grow to understand our problems and their solutions better.