KEVIN VAN HORN 258 Thornberry Way Pleasant Grove, Utah 385-439-7784 kevin@ksvanhorn.com SENIOR BUSINESS ANALYTICS ENGINEER Business Analytics Technology, Technical Analysis, Software Design, Development, and Implementation Signature Strengths * Solutions Development/Implementation * Analytics & Modeling Methodologies * Best Practices & Quality Control * Competitive Intelligence * Project Management * Performance Management/Improvement * Staff Leadership/Development * Experimental Design * Accomplished, results-driven information technology specialist with a broad-based background in the development and implementation of business analytics technology and experimental design software. * Proven experience identifying, developing, and implementing technical solutions to enhance competitive market positioning; well-versed and hands-on expertise in applying Bayesian statistical and Markov Chain Monte Carlo (MCMC) models and methodologies. * Regarded for the ability to lead framework development and all phases of the SDLC with success in motivating cross-functional teams, implementing best practices, and achieving business, technical, and quality goals. * Works well under pressure to manage and meet aggressive deployment deadlines. Objective and progressive with dynamic leadership, technical, and business acumen. PROFESSIONAL EXPERIENCE THE MODELLERS, Midvale, Utah [2009 – 2015] Principal Engineer Provide senior level guidance and advice on the development and implementation business analytics technology, including the use of predictive analytic models to achieve organizational and client goals. Oversee the development of experimental design software, enabling flexible design restrictions to position the company as one of the only firms capable of handling complex design restrictions. Key Contributions: * Improved software performance for the Interactive Decision Tool by profiling the multinomial logit model code to identify bottlenecks and speed the attribute lookup process by a factor of 4.5. * Rescued a project to implement a new source-of-volume choice model based on an academic paper. - Analyzed R code received from the author and led efforts to re-implement the model in C# by creating a set of unit tests to identify issues. - Employed MCMC expertise to uncover problems within the C# code and issues in the original R code. - Developed a new technique for Markov chain Monte Carlo when the likelihood is a Monte Carlo estimate. * Developed and maintained internally used software to handle various forms of analytics. - Proposed the mathematical model, derived MCMC updates, and implemented estimation and prediction. - Worked out a variant of the source-of-volume model called Similarity Matrix Probit which does not assume a predefined product similarity function but infers similarity function weights from survey data. * Led efforts to improve software that estimated Bayesian hierarchical multinomial logit choice models, including implementing convergence diagnostics, parallelizing implementation to allow multiple processor cores, improving MCMC updates enabling better mixing of the Markov chain, implementing an initial step to estimate a simplified version of the model to allow early warnings of data problems, and designing facilities to checkpoint computation state to allow the estimation process to continue if a machine crashed during a long estimation run. * Implemented a non-compensatory choice model variation called conjunctive screening based on the theory must-have features or thresholds for feature values must be met prior to an individual making a product purchase. * Designed and implemented an experimental design capability for multinomial logit consumer choice models, employing a Bayesian D-optimality criterion, enabling the user to specify a broad-range of design restrictions. * Identified and resolved issues where the predictions of models estimated from survey data did not match calibration task numbers for “would you buy?” questions due to deficiencies of the mathematical model itself and use of the posterior mean point estimate. - Developed a method to determine optimal point estimates to produce choice probabilities to closely match Bayesian posterior predictive distribution. - Modified the choice model to use a restricted form of nested logit that groups the “none” option separately from the product options. * Enhanced a proprietary application and improved survey data analytic capabilities by adding a Bayesian pathways model, a regression structure designed to model the stages of the marketing funnel, enabling the ability to distinguish directs effects from total effects of a variable on the final outcome, corrections for individual differences in scale usage, and performing automatic variable selection. * Provided guidance and advice in advanced statistical methodology, supporting the successful acquisition of a $4.5 million tracking study contract with Microsoft results to determine effectiveness of marketing campaigns; secured contract approval by key Microsoft decision makers by incorporating a Bayesian state space model to contain costs, maximize survey result liability, and allow smaller sample sizes. * Drafted a proposal and delivered to Consumer Reports on methods to improve analysis of product reliability surveys via the implementation of hierarchical Bayesian Poisson process model to track breakdown rates by brand and other variables. - Utilized findings to compute the probability of a severe breakdown within a given time after purchase. - Developed the mathematical model, analyzed a set of data, and delivered an entire software package and detailed manual to enable the company to conduct analysis on future data sets. THE GENERATIONS NETWORK, Provo, Utah [2008 – 2009] Senior Development Architect Recruited to lead the extension of search capabilities of genealogical search services. Spearheaded the creation and analysis of a queuing theory model designed to evaluate required computational resources and cut search hardware service costs. VINTELA INC. / QUEST SOFTWARE INC., Lindon, Utah [2004 – 2008] Product Architect and Team Lead Oversaw efforts to extend Microsoft’s Operations Manager and Configuration Manager products to effectively administer Linux and UNIX systems. Key Contributions: * Designed and implemented a privilege separation monitor to enhance the security of the QMX line. * Appointed team or co-team lead for several QMX products, holding full accountability for all aspects of the Scrum development process. * Directed the launch of a comprehensive quality improvement program, encompassing code reviews, team mentoring, test-driven development, and the use of mock objects for unit testing and educating developers. * Drove efforts to meet all project milestones by participating in Scrum sprint planning sessions to ensure on-time and within budget completion. * Mentored and trained developers to ensure an understanding of best practices, processes, and protocols; assisted in identifying and resolving complex issues. * Championed the successful turnaround of a VSM project by introducing the use of test-driven development, unit testing, and code reviews, enhancing overall product quality. CAVEON LLC, Draper, Utah [2003 - 2004] Programmer, Contract Provided consultative guidance and advice to support the development and delivery of the WebPatrol application. Key Contributions: * Automated the process of search for and identify high-stakes exam fraud and compromise on the Web by drafting specifications, designing, and implementing Caveon’s WebPatrol system. * Utilized Java, MySQL relationship database, Lucene text indexing and search engine, and pattern recognition as a page classifier designed to recognize relevant Web pages. NORTH DAKOTA STATE UNIVERSITY, Fargo, North Dakota [2000 - 2003] Assistant Professor, Department of Computer Science Conducted research on applying the method of maximum entropy to acoustic model for speech recognition. Key Contributions: * Invited by the Institute for Signal and Information Processing to review the software design of the speech-recognition toolkit; reviewed the design and application programming interface (API) to identify areas for improvement. * Developed and delivered a graduate-level course in generic programming and C++ library design. Early Career History: Senior Developer and Technologist, Tenfold Corporation; Research Scientist, Fonix Corporation; and Software Engineer, Excite Inc. EDUCATION BRIGHAM YOUNG UNIVERSITY – Provo, Utah Ph.D. in Computer Science Awarded: Computer Science Departmental Ph.D. Fellowship CALIFORNIA INSTITUTE OF TECHNOLOGY – Pasadena, California Master of Science in Computer Science Awarded: AT&T Bell Labs Scholarship and Fuller Fellowship UNIVERSITY OF UTAH – Salt Lake City, Utah Bachelor of Science in Physics and Computer Science (Dual) | Graduated Magna cum Laude Awarded: National Merit and President’s Scholarship Member: Phi Beta Kappa Honors Society TECHNICAL EXPERTISE Applications Software: JAGS; WinBUGS; LyX; LaTeX; and Microsoft PowerPoint, Excel, and Word Software Development: Stan; Rstudio; Visual Studio; Subversion; Git; Nunit; GCC; Make; and emacs Programming Languages: C#; C++; C; R; Python; Perl; Common Lisp; Haskell; Scheme; Java; Pascal; HTML; and SQL Operating Systems: Linux; UNIX; Mac OS X; and Windows Models & Methodologies: Bayesian Statistics; Markov Chain Monte Carlo Methods; Discrete Choice Models; Multilevel/Hierarchical Models; and Ordinal Regression RECOMMENDATION "Kevin is the best developer I’ve ever met with respect to delivering defect free software. We worked together on our internal Sherpa decision modeling software and in six years, I can’t recall a single defect in his code that had any impact on customer projects or the credibility of our development team or its software. Not surprisingly, he also is highly committed to automated unit testing. If all developers were as committed to quality and unit testing as Kevin, the software industry’s quality problem would vanish.  In addition, his domain knowledge in the area of Bayesian analysis rivals that of the most accomplished researchers in the field. In fact, a number of times, internationally respected academics asked Kevin to review their work and manuscripts, and he was able to raise important points, find hidden issues, or provide better approaches that added to the quality of their work. In other cases, his innovative modeling ideas were the difference between winning and losing projects for us. It is very unusual that a developer has that kind of impact at The Modellers. With that domain knowledge, naturally Kevin also successfully delivered the most complex computational features in our choice modeling software. Kevin is easy to get along with and genuinely concerned about the success of his team’s effort. It was a pleasure to have him report to me, and I enthusiastically recommend him for any senior developer position." David Guell, Sr. VP of Software Engineering at The Modellers