Mark Stevenson

• Built Credit Risk model to ID low-risk bidders
• Built Item-Based recommendation engine
• Supervised construction of Gross Merch. Value model
• Used Selenium for open-source data aggregation
• Implemented fuzz-matching routine to normalize address data
• Set up Google Analytics
• ID'd operational inefficiencies & opportunities for improvement
• Provided ad-hoc statistical analyses.

• Leverage Team Strengths to Algorithmically Identify Emerging Threats within IRS
• General Software Development, to Include Maintenance, Debugging and Expansion of Existing Code-base
• Develop ad-hoc use-case Scripts
• Develop Anomaly Detection Models
• Develop Graph/ Network Models
• Document Models
• Validate Models
• Interpret and Brief Models
• Model Expansion
• Use-Case Development
• Data Sourcing
• Database Development
• Participate in bi-weekly Client Briefings & Status Updates
• Generate Monthly Threat Reports
• Track General Administrative Actions, i.e. Pending Clearances and Onboarding Statuses

Model Development/ Algorithm Implementation

• Developed Unsupervised Deep Learning Model to Detect Behavioral, Network anomalies.
• Developed Tree-Based Models for Risk Classification.
• Employed Bayesian Optimization (Hyperopt) to optimize DL models in a timely manner.
• Implemented Levenshtein Algorithm (NLP), Regex to identify unauthorized email transmissions.
• Used PCA, Kmeans to ID Anomalous Network Activity.
• Used GAM to identify time-series anomalies in user behavior.
• Developed Dashboard app to visualize anomalous network activity
• Developed Java Routine to Expedite Completion, Dissemination of IRS Audit Documents.

Statistical Projects

• Worked with DoD to design behavior modification experiment.
• The DoD wanted to determine whether behavior can be modified to save energy in buildings.
• Given different categories of buildings with different attributes, what are the minimum populations necessary to conduct experiment with a given level of statistical certainty?

ETL Development

• Perform ETL on financial, behavioral, and network data in order to ID emerging threats within the IRS
• Built dozens of risk metric classes to ID anomalous activity within the IRS
• Classes are attached to different datasets
• Due to inconsistencies across the production environment, multiple languages had to be used
• Process driven by a dynamic Bash driver
• Each script produced a table containing key, and one or more risk metrics
• Tables recursively joined to arrive at a machine-learning ready dataset

Software Engineering

• Developed Dashboard app to visualize anomalous network activity
• Parallelization of long-running scripts
• Saved client 20+ hours/ week by automating workflow processes.
• Developed LiveCycle Filling Java Routine to Expedite Completion, Dissemination of IRS Audit Documents.

• Build Pareto-Optimized Portfolio of Sensor Configurations for Camp Humphreys, South Korea
• Project Involved a Combination of Systems Engineering and Plume Modeling.
• Tools Utilized: R, VBA

• Model Documentation Review
• Code Review
• Conceptual Review
• Replication of Results
• Performance Measurement, Benchmarking
• Stress Testing
• Consultation with Model Owners, Stakeholders
• Tools Utilized: R, SAS, VBA.
• Reverse engineering of Bayesian forecasting models for testing purposes
• Developed Automated Schedule Management System

• Developed & Implemented Analytics Packages to Assess Impact of Marketing Efforts
• Leveraged Python, SQL and SAAS to Automate and Streamline Existing Procedures
• Conduct ad-hoc Statistical Analysis as Needed
• Supervised Subordinate Analysts, DBA

• Leverage 26 troops to fund and account for various civil, military efforts in northern Afghanistan.
• Expanded financial management operations from 1 outpost to 3 permanent offices.
• Worked with the Boston Federal Reserve to stand up EagleCash system in area of operations.

• Ensure staff sections are equipped and coordinated.
• Responsible for maintenance and accountability of over $1 million in vehicles and equipment.
• Responsible for establishing initial contact with client units; primary point of contact during training.
• Similar to project management.