Resume indeed

I have a broad background in automation of expert-level processes, with specific training and expertise in key research areas of discovery, validation and commercialization for this application.


During my Ph.D. work at Los Alamos National Laboratory, I developed a mixture-of-experts software process for interpreting contaminated and mixed chemical samples that did not look like their expected, clean training standards. I worked with chemists and a set of pattern recognition methods to determine the strengths and weaknesses of each method. I implemented and delivered software that recognized the presence of confounding scenarios and automatically reassigned the weight on each method based on its past performance in the scenario. As a result, I produced a method that reported an automated quality score for each result as well as a folio of the evidence discovered through the processing steps.


At the laboratory automation provider Coyote Software, I wrote software to observe, learn and interact with a real-life process being developed ad-hoc by an automated laboratory. I developed an algorithm that watched for and timed delays in robotic movement in order to do as many operations as possible without missing hard incubation deadlines because of a delay. This artificial intelligence based program outperformed the industry standard method of static scheduling for both well-behaved operations and for errors that are inevitable in a system with mechanical parts.


At the pharmacoinformatics company Bioreason, I applied a mixture-of-experts approach hierarchically to the task of evaluating candidates for agricultural pesticide and herbicide from hundreds of thousands of theoretically possible compounds. Like other expert-level processes, the protocol for this process was not exact and the experts could report differing yet correct results. I captured the fuzzy profile of the protocol by observing and tracking the decisions made by experienced senior chemists, and found correlated sources of quantifiable information that could be developed about a compound without human expert intervention. I implemented and delivered the resulting mixture-of-experts system for use at a major Agricultural company.


At MitoTech LLC I led the algorithm discovery and development work to automate the forensic analysis of mtDNA sequences. When again faced with an expert-level protocol that described the intention of the process but not each step that could be encountered, I again used data mining techniques on the data and interviewed and observed the humans to discover the actual process preferred by the human experts. After validation of the recodified algorithms by the Federal Bureau of Investigation, they were implemented and delivered to the FBI to continue to improve the justice and missing persons processes.

I bring a deep, broad background in process analysis, automation and commercially-relevant production that began at the ground-level as a professional programmer supporting a 24/7 AT&T manufacturing operation. Delay or inaccuracy translated instantly to financial loss. In addition to expanding to roles as technical and project leader, and database administrator, I tested for and obtained certification in the domain of process science itself through the American Society for Production and Inventory Control (APICS). These experiences gave me an eye to look for the underlying process, the factors and variances that are significant and gave me the basic skills needed to hunt down the elements that are key to successfully automating part or all of a process. My Master of Science studies in theoretical mathematics added domain translation to these rules, since theoretical mathematics is a study in mapping, comparing and contrasting between various models. It led me to develop rules and expert sense about when a feature can be assumed to be tracking with something else and when it has to be verified on its own. Further computer science studies in RISC, operating system process and resource management, and compiler language design and optimization all added layers to my ability to watch and analyze a process. My Ph.D. work was in the artificial intelligence field as applied to expert-level processes.


Publications, Patents and Commercialized Software Produced

Publications:

o        Den Hartog, B.K, Elling, J.W. & Budowle, B. (2009) The impact of jumping alignments on mtDNA population analysis and database searching. Forensic Science International: Genetics Supplement Series - December 2009 (Vol. 2, Issue 1, Pages 315-316)
o       Den Hartog, B.K, Elling, J.W. & Kepler, R. B. (2009) MTexpert, an automated software system for forensic mitochondrial DNA data analysis. Forensic Science International: Genetics Supplement Series - December 2009 (Vol. 2, Issue 1, Pages 102-103)
o       Budowle, B., Fisher, C.L., Polanskey, D., Den Hartog, B.K., Kepler, R.B. & Elling, J.W. (2008) Stabilizing mtDNA sequence nomenclature with an operationally efficient approach. Forensic Science International: Genetics Supplement Series - August 2008 (Vol. 1, Issue 1, Pages 671-673)
o       Budowle, B., Fisher, C.L., Polanskey, D., Den Hartog, B.K., Kepler, R.B. & Elling, J.W. (2008)  Software program for standardizing an operational mtDNA nomenclature for forensic analyses, DNA In Forensics, Ancona, Italy, May 30, 2008
o       Budowle, B., Fisher, C.L., Polanskey, D., Den Hartog, B.K., Kepler, R.B. & Elling, J.W. (2007)  “Standardizing the Nomenclature for mtDNA Haplotypes with an Intuitive Hierarchal Execution Software Program”, International Society for Forensic Genetics,  Copenhagen, Denmark
o       Bacha, P.A., Gruver, H.S., Den Hartog, B.K., Tamura, S.Y. & Nutt R.F. (2002)  Rule Extraction from a Mutagenicity Data Set Using Adaptively Grown Phylogenetic-like Trees, Patricia Nutt  J Chem Inf Comput Sci 42(5): 1104-1111  
o       Den Hartog, B.K., Elling, J.W. &  Kieckhafer, R.M. (1999) The SI-Combiner: Making Sense of Results from Multiple Trained Programs.  Lab Automation ‘99, San Diego, CA.
o       Den Hartog, B.K. (1998) A Fuzzy Logic Framework for Integrating Multiple Learned Models. Ph.D. Dissertation, University of Nebraska at Lincoln.
o       Medvick, P.A. & Den Hartog, B.K. (1997) A Script-Directed Controller of Modular Automation (SDCMA) for Genome Laboratories. Book chapter in Automated Technologies for Genome Characterization. Editor T. J. Beugelsdijk, John Wiley & Sons, Inc.
o       Den Hartog, B.K., Elling, J.W., Mniszewski, S.M. &  Kieckhafer, R.M.(1996) Learning from Multiple Models to Integrate Sensed Data. Proc. AAAI Workshop on Integrating Multiple Learned Models for Improving Machine Learning Algorithms, Portland, OR.
Patents Pending and Commercial Software Produced
 
o       Inventor, U.S. Patent Pending, Computer System and Computer-Facilitated Method for Nucleic Acid Sequence Alignment and Analysis, 2009 Patent Application Number 20090292665
o       Software Release MitoTyper™, Quantico, VA, 2009; Forensic mtDNA sequence profiler
o       Software Release MTExpert™, Quantico, VA, 2009; Forensic mtDNA sequence expert system
o       Software release ClassPharmer™ (2003); Clustering, data mining and visualization for lead identification
o       Software release LeadPharmer™ (2001); Clustering, data mining and dashboard for lead identification
o       Software Release Coyote Controller (1999), Hamden CT; Learning controller for robotic laboratory

Subscribe to: Comments (Atom)