Andrew Jones

• For a scaling specialist insurer focusing on cybersecurity offerings, leading the process design and model development for an autonomous underwriting system that can profitably quote core products without human intervention

• Lead Data Scientist and Product Owner for the team that developed a core underpinning system at CFC. This system enabled correct firm identification and entity resolved a huge variety of attributes to be used for automation, pricing, risk selection and cross sell purposes throughout the business

• Developed a document parsing system that extracted attributes needed for automation decisions. The first impact of this system is how it's enabled CFCs renewal book to proceed autonomously at policy expiry.

•For a leading digital insurer, led a variety of projects focused on: market average price prediction, large loss prediction, claims frequency prediction, modernizing deployment architecture via cloud services (Azure Databricks) and establishing MLOps processes.

• Developed and productionized two market data based XGBoost price prediction algorithms to significantly increase sales and average policy premium. Effort involved all typical aspects of data science model development including data sourcing through model selection & hyperparameter tuning.

• Directed a small team in developing and operationalizing an MLOps platform on Databricks. This solution enhances agility by enabling a balance between model complexity and response time which is crucial for accommodating the quick response time cutoffs imposed by aggregator sites.

• Created model interpretability modules using SHAP values for increasing trust in an NLP based claims prediction algorithm.

• For a major DoD customer, engaged in creating various data pipelines using Python to connect disparate source systems data before developing Tableau views. These views will be implemented as part of a governance process to reform readiness understanding within the organization.

• Using open source AutoML technology and working with engine experts, created a blade failure algorithm to correctly predict failures in this critical engine component. Involved typical Data Science steps of data exploration, feature engineering (to include creating synthetic records), model selection, evaluation and user acceptance.

• Using Docker, established the correct dependencies and containerization procedures to allow a critical engine health algorithm to deploy in any technical environment. This is expected to be utilized across several rotorcraft platforms, become a key component of predicting engine health, and improve predictability at many downstream supply chain and maintenance organizations.

Engaged as an independent consultant to provide system architecture and pricing for a new commercial data platform. Pricing project involved the creation of a Monte Carlo simulation to account for unknown variables & provide a holistic Total Cost of Ownership (TCO) to the organization. After the successful delivery of this project, the client is continuing to move forward in constructing the platform we designed & priced.

• Diverse role including business development, project leadership, and solution development for a small Data Science consulting firm with a focus on Machine Learning based solutions.

• Led the successful pursuit of several Machine Learning focused projects for various firms, including Fortune 500 companies. This process often involved many phases such as use case development, solution ideation, technical proposal writing & delivery, and proof-of-concept development.

• Took the lead functional role in the delivery of several diverse cutting-edge projects such as Deep Learning for Time Series Forecasting, translating natural language to SQL (NLIDB), and NLP for pharmacovigilance case processing

• In response to a maturing marketing platform, pioneered a suite of R scripts that resulted in major efficiency gains by automatically processing millions of records from hundreds of files into various reporting mechanisms and 200+ metrics.

• In preparation for the 2020 census, led a statistical analysis that combined several demographic factors into a linear regression model for consumer response rates within 39,000+ distinct ZIP codes.

• Successfully positioned our team to win future work by designing and socializing an executive dashboard in Qlik Sense that visualizes system performance, user interaction, and response rates across the United States.

• Created the foundation for a pricing mechanism by developing a Monte Carlo Simulation model that predicts campaign level consumer response rates from various input factors.

• For high visibility IT projects costing over $5 million dollars, led the process and junior resources that would price around $2 billion of IT investments. The project financial estimates would be used to inform an investment committee's decisions for VP level IT and finance clients

• For a budget of approximately a billion dollars, implemented a Time Series Forecasting tool that accurately predicted IT spend to within a 7% error and informed critical funding re-allocations