Cloud Computing and Analytics

Lowe

Andre Lowe

Intel Corporation

Cloud-based Valuation Tools at Intel Corporation

Corporate investment decision processes that leverage portfolio theory are starting to become popular for both R&D and Capital activities. While there are many commercially available software packages to help decision-makers visualize the ROI implications of their choices, a consistent short-coming of these packages is that they presume the incremental value of each decision is a known input for the tool. To help ensure consistent, high-quality valuations associated with business portfolio-level choices, Intel has set about creating a suite of tools. In this presentation, Mr. Lowe will discuss the business modeling and valuation aspects of the tool with particular focus on the conceptual structure of the model, acceptable short-cuts and simplifications, overcoming cultural resistance and communicating the results during different steps in the funding process.

Bio

Mr. Lowe has been with Intel Corporation for 10 years in various roles in finance and decision support. Most recently, Mr. Lowe’s focus has been on incremental business valuation and investment portfolio optimization to support Intel’s semi-annual funding cycles.

Back to schedule

Megahed

Aly Megahed

IBM Research

Service Design Analytics for IT Service Providers

Large IT service providers compete to win highly-valued outsourcing IT deals via designing and submitting proposals to potential clients. In this talk, you will learn about the kind of analytics, machine learning, and OR work done for managing such complex service engagements. A case management approach that analyzes costs and prices of deals in preparation will be presented.
The competitiveness of deals are assessed based on mining comparable historical peer deals. After submitting the proposals to clients, many factors, beside the price, can affect the chances of winning a deal. A predictive analytics tool, based on combining quantitative and text analytics models, to identify these factors and prioritize all deals in the pipeline was developed. The tool can help sales executives bring excelling engagements to a faster close as well as identify troubled ones. This tool will be presented along with the lessons learned from using this overall
methodology.

Bio

Dr. Aly Megahed is a research staff member at IBM’s Almaden Research Center. In his current job, he develops analytical tools for complex service engagements and advances research in analytics, statistics, machine learning, and operations research to address different service science problems. Dr. Megahed got his Ph.D. in Industrial Engineering from Georgia Tech with a focus on the development of operations research and analytics tools for solving problems in supply chains and logistics systems. He has two master’s degrees in Industrial and Production Engineering from Georgia Tech and Alexandria University, respectively, and a B.S. in Production Engineering from Alexandria University. He has done multiple analytical research/consultancy projects for over 6 companies in the past, has given talks at several conferences, companies, and institutions/universities, and has his work published in several academic journals and conferences.

Back to schedule

Thomas Olavson

Google

O.R. Case Studies in Cloud Infrastructure Planning at Google

Google’s Cloud infrastructure consists of the servers, network, datacenters and software that run all of Google’s services like Search, Gmail and Youtube as well as it’s fast growing external business, Google Cloud Platform. I’ll discuss case studies in optimizing this infrastructure, covering challenges not only in solving the analytical problems but also in solving the business problems to reach real implementation and impact:

  • How do we optimize buffers for long leadtime datacenter capacity, aligning with stakeholder expectations and coordinating across a complex portfolio of construction projects?
  • What are stranded compute and storage resources and how do we measure and manage them?
  • How can we increase compute utilization through oversubscription, overcoming both technical and cultural challenges?

Bio

Thomas Olavson is director of the Operations Decision Support group at Google. His team provides model-based decision support for Google’s cloud infrastructure and supply chain planning. He was previously director of HP’s Strategic Planning and Modeling team, an INFORMS Prize and Edelman Award winning team. Thomas received his Ph.D. in Management Science & Engineering from Stanford University.

Back to schedule

VanDenHeever

Susara van den Heever

IBM

How to Leverage Optimization in a New Generation of User-centric Analytics Platforms

Even though many Line-of-Business users emphasize the importance of analytics, several of them identify significant obstacles in applying analytics effectively, such as complexity of available tools, limited access to data, and limited time. Recent advances in cloud and cognitive computing made possible a new generation of analytics platforms which focus on delighting users and simplifying collaboration, without requiring in-depth knowledge of underlying analytics techniques.
In this talk, you will learn:

  • How we envision optimization as part of this user-centric analytics experience, together with descriptive and predictive analytics.
  • How advances in natural language processing, cognitive computing, visual tools, and mobile and social sharing combine for a more delightful optimization experience.
  • How these new platforms could expose self-serve optimization to a wider range of business functions.
  • Through a case study involving a sales manager, how to use these platforms to tell a compelling end-to-end analytics story, highlighting the benefits of optimization.

Bio

Dr. Susara van den Heever is the Product Manager for IBM Decision Optimization on Cloud and the Program Manager for Joint Projects between IBM Research and IBM Decision Optimization (ILOG). She is passionate about making analytics more understandable, consumable, and enjoyable to the fast-growing user group of descriptive, predictive, prescriptive analytics. Previously, she managed the Constrained Resources and Environmental Analytics (CREA) team at the IBM Smarter Cities Technology Center at IBM Research Ireland. Her interests include sustainable optimization-based solutions to design and planning problems under uncertainty for a variety of industries, including energy and utilities, oil and gas, and consumer goods. Susara holds a Ph.D. from the Department of Chemical Engineering at Carnegie Mellon University, and has over 18 years of experience in research, consulting, management, and education for optimization-based industry solutions.

Back to schedule

Zhenyu Zhao

Yahoo

A Test of Minimum Difference for Online Controlled Experiment with Real Examples

Online controlled experimentation (A/B testing) is widely used for evaluating innovative product development at Internet companies, for both frontend design and backend algorithms. Conventional statistical tests are able to evaluate if the difference between two experimental groups is different than zero. However, experiment owners are sometimes interested in studying if the difference between the two groups is of greater than a certain magnitude. Using conventional tests therefore falls short of providing the answer to this question. A novel statistical test method is proposed to determine whether the observed difference is greater than a minimum pre-specified magnitude. Mathematical formula and examples will be included to showcase how we implement the proposed method in practice.

Bio

Zhenyu received his PhD from Northwestern University in Statistics, and started as a Sr. data insights analyst at Yahoo after graduation.  Zhenyu has hands-on experience in online experimentation for multiple Yahoo products. He also contributes to the development of online experimentation platform. For online experimentation, he developed several novel approaches to tackling practical challenges and satisfying business needs. He also conducts ad-hoc analysis using advanced statistical and machine learning model to answer key questions.

Back to schedule