D A T A S C I E N T I S T

Ken Chumley

www.linkedin.com/in/kenneth-chumley/

adgywi@r.postjobfree.com

Jacksonville Beach, FL

ADDRESS

• 30+ Years Data Scientist

• Artifical Intelligence

• Realtime Machine Learning

• Neural Networks

• Deep Belief

• Regressions

• Logistic

• Random Forest

• Isolation Forest

• NLP (nltk, spaCy)

• Collaborative Filtering

• Classification

• Cluster Analysis

• Principle Components

ABOUT ME

SKILLS

Data Modeling

Data Mining

Python

.NET

C#/C++

MS SQL

WORK

July 2020

Sept 2018

South Carolina Dept of Health and Human

Services, Columbia SC

Data Scientist: capacity modeling, anomaly detection, probability density, NLP, Classification, Clustering Python, SAS, MS SQL, pySpark, Piplines, C#/.NET

Present

Oct 2017

Cetrest Corp, Jacksonville FL

Data Scientist, Classification, predictive modeling of FBI UCR data for research between crime and

governement programs in SC

Python, C#/.NET, MVC, Angular, MS SQL

HOBBIES

Boat building, USCG Master Captain, metal working, wood working, astronomy, robotics, writing, prepping, hiking, backpacking, camping 302-***-****

Masters of Science in Computer Science

KWU/Georgia State University, Adaptive Analytics

Oct 2017

May 2015

Cetrest Corp, Jacksonville FL

Data Scientist, Created Real-time adaptive analytical system called ARTA

Python, C#/.NET, C++, MVC, MPF, MS SQL

EDUCATION

BS Natural Resources

University of Georgia, concentrations in Computer Science and Operations Research, with post graduate studes in statistics, finance and management.

Masters of Science in Biometrics and Applied Operations Research Cooperative studies between University of Georgia & Meade Corporation.

KENNETH CHUMLEY RESUME BIOGRAPHICAL

SUMMARY

• 30+ years as data scientist

• 20+ years as programmer/system analyst

• 15 years as project manager

• Strong expertise in modeling, real-time analytics, data minding, real-time Microsoft Windows applications for device control, medical technology and interactive graphics

• Available for full and part time assignments. adgywi@r.postjobfree.com or 302-***-**** EXPERIENCE

South Carolina Dept of Health and Human Services, Columbia, SC 09/2018 to 07/2020

Data Scientist

Provided advanced data science services of medical records for various needs including capacity modeling, visualization, data cleansing, fraud analysis, probability density, classification, cluster analysis, database architecture, natural language processing, data mining and complex SQL coding. All analytical work was done using various Python libraries to support data-driven decision support needs. SAS, SQL and various web sites were used for data mining.

AI/ Machine Learning was applied in supervised data environments using claims data from both Medicaid and Medicare sources. Regression/Classification approaches included K Neighbors Classifier, Linear SVM

(Support Vector Machine), RBG SVM (Radial Basis Function), Gaussian, Logistic, Decision Tree, Random Forest, Naïve Bayes Gaussian, Neural Networks, AdaBoost, CatBoost, Autocoder and Quadratic Discriminant Analysis. GridSearchCV was used to discover optimal parameters and the Voting Classifier for model assembly.

Clustering of unsupervised data included several approaches, including K-Means, Mean-Shift, Agglomerative Hierarchical and Affinity Propagation to analyze changes in drug prescriptions after the start of the COVID-19 Pandemic. Segmentation used to identify key features and associated affects in fraud modeling.

A Principle Components Analysis (PCA) was used in high dimensional datasets for the pre-correlation of inputs. Isolation Forest was applied to find outliers for fraud identification and data cleansing. Contour plotting was used for visualization. These approaches were used to identify outliers from normal patterns of drug prescribing by providers within the state.

Receiver Operating Characteristic (ROC), Precision, Recall, TPR/FPR and Confusion Matrix used to judge quality of models and bias control.

NLP of provider notes and deep learning analysis were used to classify various treatment plans and to recognize signification details. Significant verbs, bigrams and general settlement were employed to facility classification. Advanced features of SpaCy were used including prediction of syntactic dependencies, pattern matching, semantic comparisons, pipelines and custom attribute/property/method extensions. The genism library was used for analysis of word vectors using TF-IDF coding. pySpark distributed processing was used when single processor time exceeded 24 hours. Environment: Python (Scikit-learn, TensorFlow, Keras, inflection, fuzzywuzzy, jupyter-client, seaborn, pyodbc, scipy, warlock, nltk, spaCy, genism, pySpark and various common libraries), C#/.NET, MVC, Angular, MS SQL, Cassandra NoSQL and SAS

Cetrest Corporation, Jacksonville, FL

10/2017 - Present

Data Scientist

Provided advanced crime analytics using the FBI-based Uniform Crime Reporting data (UCR) for research into relationships between crime and government programs such as the SC Dept of Health and Human Service Medicaid programs and Medicare from CMS. Designed a pattern recognition system for classification and predictive modeling of crime with risk assessment for generic or human profiles. The risk factor was used as a labeled outcome and analyzed using various forms of classification/regression machine learning including Regression/Classification approaches included K Neighbors Classifier, Linear SVM

(Support Vector Machine), RBG SVM (Radial Basis Function), Gaussian, Logistic, Decision Tree, Random Forest, Naïve Bayes Gaussian, Neural Networks, AdaBoost, CatBoost, Autocoder and Quadratic Discriminant Analysis. K Means Clustering was used identify similarities between offenders for possible associations.

Environment: Python (Scikit-learn, xgboost, lightgbm, catboost, seaborn, pyodbc, scipy and various common libraries), C#/.NET, Angular, MVC, MS SQL

Cetrest Corporation/ARTA Solutions, Jacksonville, FL 05/2015 - Present

Data Scientist /Project Manager/Architect/Statistician Designed and implemented an adaptive real-time analytics product/service called ARTA. The system provides horizontal elasticity and real time data processing from distributed data collection hubs, and real time updates on predefined or adaptive model parameters. The time required for model updates is independent of the number of rows, with the same amount of time required with 100 record and 10 billion records data sources. The Machine Learning Algorithms Linear, Nonlinear, Optimization, Bayesian, Random Forest, Neural Networks, Deep Learning, Deep Belief, Support Vector Machines, Collaborative Filtering, K Nearest Neighbors, Learning Vector Quantization, Bag of Words and Natural Language Processing are supported in a supervised and non-supervised environment. Data mining is implemented using an internal spider for web crawling of targeted domains. Parallel node processing supported with best practices threaded/task implementations. Hyperparameters Optimization used for initial and dynamic model structures.

Most of ARTA was originally written as a Fortran application and later revised into modern .NET languages (C#, Web APIs) with some highly optimized routines in C++. ARTA has been used in multiple data analysis roles in private and public environments. The system links to cloud services via AWS and supports remote connectivity using Azure. SAS conversion (commands and SQL code) is supported. Recent implementations include ARPC and Checkpoint. At ARPC, ARTA was able to reduce processing time of a complex analysis from 5 minutes per iteration down to 5 milliseconds. At Checkpoint, ARTA was able to remove an analytical bottleneck to allow a distributed data intelligence collection strategy with modeling at a central location.

Environment: C#/.NET, C++, MVC, Angular, ASP, R, Open R, R Studio, Python (Scikit-learn, Keras, Tensorflow), NLP (nltk, spaCy), WCF, WPF, WCF, Entity Data Models, MS SQL Server 2016 (R Services), Win Forms, Web Forms, ReactJS.NET, Azure, AWS (Amazon Web Services) Virtual Cloud, Cassandra NoSQL, Angular JS, JavaScript, JQuery, AJAX ECom Solutions, Jacksonville, FL

05/03/2017 – 04/06/2018

Data Scientist, Project Manager, DBA, Architect, Lead Developer Architected solutions and managed a small team of developers and QA members (7) for an e-commerce web application handling product listings on Amazon including inventory control, order processing, data mining (scrapers and spiders), web hooks and cloud storage (Azure). The system uses 3 domains, 4 devoted servers and 75 services (mix of RESTful, SOA, Topshelf) to provide approximately 500 clients with Amazon store management applications. Nutrition required becoming the lead developer during the second half of the project and most of my work was centered around development of WCF services – some with hundreds of parallel processes using best practices threaded approaches and named pipes. ARTA, python and R were used to perform machine learning analysis on historical product rankings (80 million records).

Created an industry specific spider (web crawler) to search the web for particular data features related to ECom needs.

Environment: Python (Scikit-learn, matplotlib, nltk, spaCy), C#/.NET, C++, AWS, MVC, WCF, JavaScript/AngularJS, RESTful services, Topshelf, MS SQL Server, IIS, S3, HTML5, Data Mining, Graphics, Analytics, ARTA, R,

Black Knight Financial, Jacksonville, FL

01/03/2017 – 03/31/2017

.NET Solutions Developer

In a short term consulting role, assisted in several of their WCF project and stored procedure needs, with some frontend development of various tools.

Environment: C#/.NET, C++, MCV, WCF, MS SQL Server, Win APIs, HTML5, Telerik, JavaScript, AJAX

Cetrest Corporation/Health Integrity, Easton, Maryland, Florida 01/2014-05/2015

Data Scientist/System Architect

Designed and implemented an automated data mining system to collect background data used in developing algorithms to identify fraudulent actions in Medicare and Medicaid. The system accessed and processed multiple federated data sources and normalized using NLP (Natural Language Processing) to assure data integrity. Machine Learning Decision Trees were used to cluster unsupervised data, which was then used in a supervised learning environment to develop non-linear and deep learning models based on actual cases of fraud identified by field agents. The system used parallel processing threads for maximum efficiency and minimum time requirements.

Environment: C#/.NET, SAS, ARTA, Python (SciPy, Scikit-learn, Tensorflow, nltk, spaCy), MVC, WCF, RESTful Web Services, Web APIs, CLEAR, WPF, MS SQL Server, SSIS, JavaScript, Razor, JQuery, AJAX, Google Services and OpenLayers

Cetrest Corporation/MTOA, Green Cove Springs, FL

02/2012-04/2016

Data Scientist/Project Manager/ Software Engineer

Developed a membership management system for the organization’s databases, statistical analysis, including payment merge from an outside source, internal money handling, member access, verification/validation and security. An OpenLayers graphical mapping system was created for visualization of supported services. The system was developed under a test-guided design/implementation methodology. SSIS was used to migrate data from an existing Access database. The tables were normalized and tuned using stored procedures. ARTA was used for statistical analysis. Developed a mobile device application to replace the existing printed roster. The mobile app provides query support against a remote SQL database and provides real-time updates of member locations on a google maps display.

I served on the Board of Directors from 2009 to present, and have held the officer positions of Membership Chairmen, Vice President and President. All work performed pro bono. Environment: C#/.NET, ARTA, Python, Apache Cordova, Firebase, MS Access 2010, MS SQL Server, SSIS, JavaScript, JQuery, AJAX and OpenLayers

Cetrest Corporation, Jacksonville, FL

10/2011-10/2012

Project Manager/System Analyst /Software Engineer

Developed an electronic charting system using bit-mapped images and vector coordinates from US Census Land based Files, and tracking using serial input from a GPS (Global Positioning System) device. The system tracked movement within boundaries of the chart system and provided various mathematical summaries of progress between target positions and planned courses. Using the combined input from the GPS, fluxgate compass and rudder indicator, the system drives a hydraulic pump capable of auto-piloting a marine vessel along a specified course. The system was multi-threaded and used real time approaches for communication and device control. A CAN thread was added for added for communication to an engine system. ARTA was used to develop a model to correct for device control error. Environment: C++/C#/.NET, ARTA, Windows 7 Real-Time, MS SQL Server, RISE UML Editor Cetrest Corporation/Siemens Diagnostics Health Care, Glasgow, DE 04/2000-09/2011

Lead System Architect/Lead Software Engineer

Designed and Implemented an OO/COM/Windows Real Time approach for the software logic of a new generation blood chemistry analyzer. The object oriented (OO) system included an instrument level, real time system (C++) with approximate 600 threads, advanced intercommunications, a process strategy for non-blocking communications between the real time system and mid-level C# systems, a pre-processing scheduler, a real-time instrument control module, Kvaser CAN Boards, and a simulation layer that allowed desk-top development without being connected to hardware. Many advanced concepts were included in the implementation, including real-time device control in a pseudo real-time environment, thread coordination, real-time to normal priority information exchange, advanced logging (XML), advanced interactive graphics, error detection and synchronous COM control. Specific responsibilities included initial design and system layout, class creation and project construction, medical record processing, development, implementation and Windows support/training. The resulting system was capable of real time control of the blood chemistry analyzer (360 motors and 9 detectors) with one computer and less than 3 milliseconds of jitter. Environment: C++/C#/.NET, COM/DCOM, ARTA, SQL, IDL, XML, Kvaser CAN, Clearcase, Windows NT/2000/XP/Vista, MS Visio, Rational Rhapsody, UML Pad Cetrest Corporation, Atlanta, GA

01/2008-10/2010

Data Scientist/ Project Manager /Statistician

In a consulting role for a major political action organization, executed data mining and applied various statistical approaches to analyze FBI Crime summaries and state reporting for a political action organization. ARTA and SAS (Enterprise and programming) were used for the analysis and reporting. A hierarchical clustering analysis provided a tree diagram by income scale and location index. The project also included conversion of work done with SAS for support in the ARTA framework. The process included data handling, SQL and math conversions to the embedded framework.

Environment: SAS, ARTA, SQL

OTHER EXPERIENCE

Data Scientist for Precision Response Corporation, Miami, FL. Specific responsibilities included statistical analysis, trend determination and data security for decision support in their client interaction business.

Data Scientist/Architect/System Analyst for Cooperative Technologies, Atlanta, GA. Specific responsibilities included development of techniques for environmental assessment and scanned image processing (full rotation, mirroring, edge detection, pattern recognition, enhancements) in real time. Project Manager/Statistician for Meade Corporation, Dayton, OH. Provided expertise in dynamic error correction and celestial mechanics using ARTA and SAS, including Fractal Pattern recognition and automated guidance systems

Statistician/System Analyst for Bell South Mobility, Atlanta, GA. Specific responsibilities included development algorithms for optimizing loading of cell phone towers using ARTA and SAS. Data Scientist/ Project Manager/ Software Engineer for Coca-Cola USA, Atlanta, GA. Developed a statistical forecasting system to project the success of various promotional activities using ARTA. Data Scientist/ Project Manager/ Software Engineer for Coca-Cola USA, Atlanta, GA. Designed and led the development of a space management system providing interactive graphics (pattern recognition), communications, charts, and statistical analysis including a linear algebra model to optimize allocations with sales and storage space as constraints.

Project Manager/Data Scientist for Mead Corporation, Rome, GA. Developed and implemented a linear algebra model for optimized harvest schedules of a timber company with 1.2 million acres of forest land using SAS and ARTA. Also responsible for the research and development of algorithms for growth, yield and mortally of forest land using various conditions and treatments.

Contact this candidate