William Meredith Email & Phone Number

Findlay, Ohio, Us

Conducted a data-driven analysis to assess the correlation between refinery outages and subsequent price fluctuations in gasoline and diesel spot and rack markets. Utilized conditional probability techniques to quantify the likelihood of price increases following outages at specific refineries.

Chicago, Il, Us

Led the development of a Question-and-Answer system, leveraging Sentence-BERT for text encoding and FAISS for approximate nearest neighbor search. Streamlined incident resolution by identifying andcomparing closed/solved tickets to open incidents, optimizing past incident knowledge to expedite issue resolution. The solution has been approved for production.Utilized a LLM and LangChain to build a comprehensive reporting system in a sandboxed environment that compiles key findings and offers strategic recommendations to enhance the customer experience. Demonstrated the tool's capabilities using StreamLit, ensuring a user-friendly interface for stakeholder interaction. Presented updates and findings on a bi-weekly basis to project stakeholders.Provided guidance and mentoring to team members at various experience levels. Assisted a colleague in the early stages of feature engineering for their model. Additionally, I conducted code reviews and offered support in PySpark programming and data analysis for junior team members.Evaluated Service Now Predictive Intelligence platform for help ticket model, comparing its performance with custom-built models. Conducted comprehensive analysis of pros and cons to inform stakeholders' decision-making process.

Us

Used ANOVA statistical analysis and a Gradient Boosting model implementation to classify potential high-risk patients for unspecified Coagulation Defects. The results were approved by clinicians and moved into production.Developed a Random Forest classification model to detect patients who potentially may have a riskadjusted diagnosis for Atherosclerosis of Aorta. The model results were productionized and delivered to the clinicians for review. Built an NLP solution utilizing Hugging Face ClinicalBERT model to perform screening for coagulation defects using provider progress notes.Used various tests and statistics including chi- square and proportion tests to select potentiallysignificant features.Met with key business stakeholders and providers to gain clinical insight for established productionmodels and ongoing operational initiatives.Operationalized and streamlined the model automation process. All models developed are currently automated at a daily cadence.Performed onboarding duties for Junior Data Scientists which included PySpark developmentalsessions and healthcare data walkthroughs.

Us

Developed a Logistic Regression model to predict the occurrence of Peripheral Vascular Diseaseamong the WellMed Membership.Engaged with stakeholders in establishing a comprehensive plan for integrating critical patient care outcomes through the organizations existing architecture.

Us

Developed a descriptive analytics report on the various progressive stages of chronic kidneydisease. Several visualization libraries including Matplotlib, and Seaborne were used to presentthe findings.