Virtual Personal Shopping Assistants
by Manu Sharma (Shopkick)
Shopkick is the most widely used physical world shopping app. With over 8MM users and 200+ retail and brand partners, shopkick helps shoppers make better shopping decisions by providing personalized deals and content as well as rewarding users for simply visiting stores. Come to this session and find out how we are using data and algorithms to create a unique real world, location aware, personalized recommender system to bridge the online / offline gap by putting a digital overlay in the physical world.
Dr. Manu Sharma is VP of growth and Data Science at shopkick. His main interests are in the fields of user behavior, viral growth, predictive modeling, recommendation engines, collaborative filtering, and drawing compelling and fun insights from data. Prior to joining shopkick, he was Director of data science at LinkedIn where he ran a team of ~30 Data scientists focused on solving a wide variety of problems for LinkedIn and beyond. As Lead Research and Development Scientist at Cake Financial, a consumer internet startup for retail investors, he designed and implemented a portfolio recommender system and a Stock / Fund rating system. He has co-authored 6 patents, and several research papers. He holds a PhD in Chemical Physics from Princeton University and a MSc in Chemistry from the Indian Institute of Technology, New Delhi.
Recommendations and Decision Support in Agriculture
by Erik Andrejko (Climate)
A confluence of factors have converged to afford the opportunity to apply data science at large scale to agricultural production. The demand for agricultural outputs is growing and there is a need to meet this demand by utilizing increasingly mechanized precision agriculture together with the enormous data volumes collected to intelligently optimize agriculture outputs. We will consider the opportunities and challenges in building recommendation systems to augment human decision making as a means to address the world’s largest optimization problem: optimizing global food pro.
Erik Andrejko leads The Climate Corporation Data Science and Research Organization, spanning research across scientific disciplines including Climatology, producing hyper-local weather forecasts, Agronomic models, connecting hyper- local weather measurements to agronomic outcomes and Geospatial models connecting remote sensing measurements to terrestrial and atmospheric phenomenon Previously, Erik worked at several Bay Area start-ups solving large-scale statistical machine learning problems. Erik holds a B.S. in Computer Science from Arizona State University and a PhD in Mathematics from University of Wisconsin – Madison.
Blending Human Computing and Recommender Systems for Personalized Style Recommendations
by Eric Colson (StitchFix)
Machine algorithms are great for tasks that require processing of large amounts of objective and structured data. However, they have difficulty with tasks that are relatively simple for skilled humans – For example, interpreting concepts in an image, or discerning tone in language, ..etc. Yet, there is a class of problems that call for precisely the combination of these tasks. This concept of human-assisted algorithmic processing is not new. It is inherent to many processes that we are familiar with. However, there are very few systems that embrace humans and machines as two resources within a single system. Instead, they are often independent and non-collaborating agents. In this talk, we explain how a single task-processing system can be architected to use diverse resources: be they human or machine. Such a system not only better utilizes each resource, but also produces better results and gets better with experience.
Eric Colson is the Chief Algorithms Officer at Stitch Fix, where he specializes in consumer algorithms. He is also an advisor at Big Data incubator, Data Elite, and Big Data Platform provider, Mortar Data. Previously, he was VP of Data Science and Engineering at Netflix and has held analytical positions at Yahoo!, Blue Martini, Proxicom and Information Resources. He holds a B.A. in Economics, a M.S. in Information Systems, and a M.S. in Management Science and Engineering from Stanford.
Prototyping Trust: Modeling the Virtuous Cycle
by Margeigh Novotny (Microsoft)
People don’t find generic recommendations particularly valuable. Our research indicates that value appears when a user and a system collaborate in the process of finding content that maps to the taste/interest/behavioral information the user is willing to share. That is, it’s an ongoing negotiation between the user and the system, where the information shared by the user is constantly evaluated in terms of whether “something better” happens as a result. Why? People reflexively attribute “human” qualities to technologies and services, so it is no surprise that this basic expectation matches their experience of rewarding human-to-human exchanges. In human-to-human relationships,people share information progressively, based on the level of trust they have with one another. So, from a user experience perspective, the question for recommendation systems is how to build trust and foster collaboration between user and system.
Margeigh Novotny is developing intelligent services for Bing at Microsoft. Prior to joining Microsoft in 2010, she was Vice President of Strategy and Experience at MOTO where she lead a cross-disciplinary team in the development of next generation product+service+experience platforms for start-ups and Fortune 100 companies.
In 2002, Margeigh co-founded the interaction design practice at Smart Design, where she drove the development of a wide range of user-focused products from house wares, mobile devices and media servers to interfaces for automobiles, airplanes, and buildings.Novotny began her professional life as an architect with a focus on digital information and physical interaction with the environment.