Data may be the new oil, but it is worthless unless you refine it into insights using machine learning, and use it to power decisions throughout your business. In this opening keynote we unveil how leading companies are outcompeting their peers by transforming themselves from being data-driven to model-driven, by:
Every company should have a business transformation strategy if they want to navigate and succeed in the rapidly changing competitive business landscape. As new challenges crop up from a consumer driven culture within an accelerated technology field, companies are looking for innovative solutions to gain deeper insight into their customers and to invigorate their line of business. For data scientists, asking the question of what problems they are trying to solve is just the beginning. Building the future with AI, while trying to keep-the-lights-on, is a problem that most companies still face today. Find out how our teams of data scientists and engineers embrace change and failure to create a data science platform that delivers some of the most innovative AI solutions in the marketplace.
Model ethics, interpretability, and trust will be seminal issues in data science in the coming decade. This technical workshop discusses traditional and modern approaches for interpreting black box models. Additionally, we will review cutting edge research coming out of UCSF, CMU, and industry. This new research reveals holes in traditional approaches like SHAP and LIME when applied to some deep net architectures and introduces a new approach to explainable modeling where interpretability is a hyperparameter in the model building phase rather than a post-modeling exercise. We will provide step-by-step guides that practitioners can use in their work to navigate this interesting space. We will review code examples of interpretability techniques and provide notebooks for attendees to download.
A data scientist’s perspective: how Gogo leverages data from airborne devices to keep inflight wifi reliable by streamlining maintenance decisions and much more. We’ll share our experience with the ML model life cycle from the initial business problem to cost savings impact and everything in between. This talk offers a look into data-wrangling, model development, and buy-in/adoption with external teams. At each step, we review the approaches taken, sharing what worked, what didn’t and why. We’ll also discuss how successful models can have far-reaching impacts across different business units and provide a springboard for future model development.
This technical session details how supervised and unsupervised learning come together in order to recommend beer. Most American bars have at least one light beer such as Miller Lite or Coors Light on tap, but the rest of the taps, as well as refrigerator space, need to be optimized with the correct products based on customer preferences. In order to do this, data science at MillerCoors collaborates with our sales team to recommend which of our beers should be focused on during the selling process. This talk will discuss how forecasting, collaborative filtering and clustering are used together for getting out the best products at each specific outlet. You’ll also learn how data science can be used to assist sales people in working more effectively.
Machine Learning has integrated itself into almost every product we use. Unfair biases can and do leak into these models, creating the opportunity for harm. Often these unfair model biases come from the data the model is built on and from an unwise assumption by the builder that the data collected is representative of the population and objectively collected.
In this talk I’ll show you a few examples of where we’ve gone wrong before and how you can reduce or prevent unfair biases from leaking into your data products.
Determining the best organizational structure for data science teams is often debated. Should the data science team be an independent function, reporting into a head of Analytics or IT, providing support to various lines of business? Should data scientists be federated, reporting into business units and operating somewhat independently from each other? Or does a matrix organizational structure provide the best of both worlds? One approach that’s becoming increasingly common is to establish a Data Science Center of Excellence — a core team of data science experts that sets the course for a company’s data science strategy, best practices and operating procedures, and technology ecosystem. They can offer guidance to citizen data scientists, data scientists reporting directly into business functions, and other stakeholders leveraging data science across the organization. But how do you go about establishing at Data Science Center of Excellence? During this workshop, we’ll walk through organizational designs, a how-to guide for getting started, and will share best practices and lessons learned based on first-hand experience building a CoE in the field. We’ll discuss what capabilities belong in a COE; whether / how to balance deep ML expertise vs. broad analytics capabilities. Attendees will leave with an actionable plan to get started on this path.
The relationship between Data Science and IT: It’s complicated. The end goal is aligned — to help the business win through model-driven innovation. Data scientists drive innovation by building models that automate or inform business processes. IT provides and manages the technology landscape that makes it all possible. IT and Data Science need to partner on the journey to create a mutually beneficial environment — a shared space that brings together infrastructure, data and tooling to foster efficient model building, testing, validation, deployment, and monitoring. But there’s tension between fueling innovation with an open environment that embraces the most cutting edge tools, and providing a place to work that is safe, governed, cost-controlled, scalable, and compliant. In this session, our speakers will share their experiences, lessons learned, and even some battle scars as they address questions around successes and failures they’ve seen partnering with IT for Data Science programs. The goal of this engaging session is to help attendees learn from collective past experiences in order to establish clearer communication lines between Data Science and IT moving forward.
Sonny Tai is a Captain in the Marine Corps Reserves and the CEO / Co-Founder of Aegis AI, a startup that employs computer vision to automatically identify gun threats in existing security camera feeds. Sonny will talk about how his upbringing in South Africa and his experiences in the Marine Corps impacted his decision to start Aegis. He will discuss how Aegis AI overcame the lack of an adequate publicly available data set in building their initial models. Their scrappiness in collecting training data through esoteric sources and creating their own training data allowed them to build a gun recognition model that performed well enough in production to deploy with initial customers and raise millions of dollars in funding from the University of Chicago and numerous venture capital firms.
Unwind from the jam-packed day with cocktails, bites, and good company of local data science leaders and practitioners.