Tenet Healthcare

AI Product Director

Trilliant Health

Developed machine learning models to predict various data points in our claims and provider datasets. My role among them was:

• working with Health Plan Price Transparency (HPPT) data to develop a proprietary Rate and Quality Score for payers and providers to enable users to identify who the high-quality *and* affordable providers are
• using machine learning to predict providers' taxonomy classification to enhance our Provider Directory product using claims and NPPES features
• using claims data to infer patients' Primary Service Area (PSA) to enhance many of our products - like our Demand Forecast, Provider Directory, and Network Explorer reaching a ~47% increase in classification accuracy

Data Scientist IV

Unum

• Developed and implemented the company’s first ever AI Governance Framework, aligning AI with enterprise compliance goals and ensuring bias mitigation, model transparency, and reproducibility
• Led the zero-to-one vision, strategy, roadmap, and execution of NLP and AI projects, building scalable AI products across different functional domains while ensuring executive buy-in
• Consistently lead user interviews to discover common pains, needs, and friction points to better inform AI projects

Principal AI Product Manager

• Lead zero-to-one AI projects through Product Discovery, Delivery, and Iteration, while using best Build-Measure-Learn practices
• Orchestrated the cross-functional launch of a scalable NLP-as-a-Service platform (backed by LLMs and GenAI) to address document classification, semantic matching, entity extraction, and summarization leading to a 54% increase over baseline models, a 73% average time savings of common manual processes which drove an annual cost savings of $2.5M
• Championed quick experimentation and iteration on AI initiatives, leading to quicker development and release cycles
• Drove executive buy-in of AI products through heavily-researched operational usability, engineering feasibility, and business viability analyses

Senior Data Scientist / AI Product Manager

Nashville Software School

Data Science Instructor - Assistant (One Cohort)

The General®

•     Led end-to-end 0-1 product development and data science execution across multiple Claims-focused ML initiatives, managing roadmaps, feature definitions, and integration with core systems
•     Identified and prioritized high-impact ML use cases in collaboration with cross-functional stakeholders, translating business needs into scalable predictive models
•     Mentored and coached junior data scientists while also setting standards for model development, engineering best practices, and product delivery
•     Designed, developed, and deployed Silhouette, an internal ML infrastructure tool that automates cloud-based environment setup (AWS), version control (DVC/S3, Git/Bitbucket), and reproducible ML workflows (Pipenv, Cookiecutter)
•     Scaled internal tooling and drove cross-team adoption of Silhouette, reducing friction in deployment and improving reproducibility across ML projects

Lead Data Scientist

•     Shaped data science team vision, product strategy, and foundational infrastructure - establishing the team’s first tech stack, creating its first public-facing AI blog, and launching its first internship program
•     Led zero-to-one development of NLP product using claim notes (Word2Vec, LDA Topic Modeling, Concept Mining) to predict attorney involvement, delivering $2.3M in business savings
•     Built and trained a deep learning model (PyTorch) to classify vehicle damage from claim images, enhancing adjuster workflows
•     Used OCR to extract structured text from scanned claim documents to identify and surface specific types of claim denials
•     Developed a REST API web app to support labeling workflows, enabling scalable image annotation for model training
•     Owned product lifecycle and modeling for a total-loss vehicle prediction system, driving $2.5M in annual cost savings

Data Scientist II

• Perform various natural language processing tasks (R/tidytext, Python/gensim), including sentiment analysis, topic modeling, and Word2Vec/Doc2Vec
• Build a nation-wide price elasticity model using a Mixed Effects/Hierarchical Logistic Regression algorithm (with cross classified and nested random effects) to predict demand for a product at various pricing structures (R, Julia)
• Predict a claim's route by converting claim notes to Word2Vec word embeddings and feeding those word vectors into machine learning models for multi-class prediction (Python/Gensim/Scikit-learn)
• Build an interactive geo-spatial smoothing app (Python/Flask, PostGIS, Leaflet.js) for various key performance indicators using AWS S3 as the data store
• Collaboratively design and implement statistical techniques and machine learning models with team members and department heads
• Create pipeline to transfer data from SQL Server database on local server to PostgreSQL database on AWS EC2 instances
• Assist the data science team in transitioning from local servers to cloud computing on AWS
• Architect and implement RESTful APIs (Python/Flask, Celery, Redis, SQL) that expose various statistical models for consumption

Data Scientist I

Perception Health (Acquired by Huron)

•	Built machine learning models (Python) that used lab data to predict if a patient would be admitted or not and visualized results using an ROC curve and confusion matrix
•	Worked alongside Perception Health's Principal Data Scientist to analyze physicians’ referral patterns and consult hospitals on how to best reduce out-of-network referrals
•	Used GIS (PostGIS, Leaflet.js) techniques to visualize healthcare trends geographically
•	Lead production on Perception Health's flagship product since March 2015 to the present
•	Built customizable, interactive dashboards (Leaflet.js, D3.js, Crossfilter.js) into the primary platform
•	Optimized SQL queries and data preparation for faster graph rendering and page-load times
•	Utilized Agile methodologies (JIRA) to organize releases while ramping up features
•	Worked with the lead architect to optimize SQL scripts across the primary product's platform
•	Converted the platform to a RESTful API for more efficiency and to decouple the server-side code from the client side templates
•	Designed and developed an internal SQL application (Python, Flask) for non-programmers to easily upload data into a database, making repetitive communication with the database faster and more memory efficient

Software Engineer Project Lead

Perception Health

• Built machine learning models (Python, scikit-learn) that used lab data to predict if a patient would be admitted or not and visualized results using an ROC curve and confusion matrix
• Worked alongside Perception Health's Principal Data Scientist to analyze physicians’ referral patterns and consult hospitals on how to best reduce out-of-network referrals

Data Scientist / Software Engineer

Relode, LLC

• Designed and wrote programs that efficiently scraped useful data from various websites using Python and created web applications to be used internally by Relode sales representatives to access this data.
• Uploaded and maintained captured data in a MySQL database.
• Wrote Python scripts that reformated and correctly structured raw, unstructured data.

Software Developer (Contractor)

Digital Reasoning

• Parsed corrupted XML files in Python using ElementTree and Regular Expressions in order to pin point the corruption
• Designed and wrote programs that efficiently cleaned and structured the corrupt data in order to increase the accuracy rate of the firm’s predictive models

Checkd.in

• Employed web-parsing techniques to convert unstructured, raw data into a workable format to perform statistical analyses
• Utilized Facebook's Social Graph API to scrape useful data
• Designed and wrote programs that parsed and statistically analyzed Facebook data using Python

Software Developer (Intern)

Building data products @ Gridsearch | Principal DS @ Trilliant Health

I partner with businesses to bring data and AI-centric products and features to market. 

I also build machine-learning apps with and for my sons. Currently, I’m building a Natural Language Understanding app I call LADNER (Locating Assignments and Dates using Named Entity Recognition) that locates homework assignments and due dates on his teachers’ websites and emails, and then automatically adds them to my Trello calendar daily so we don’t miss homework assignments or quizzes. 

Before that, I built a machine-learning app called Achoo that predicted my son's inhaler usage using air quality, weather, and allergen data.

In my career and hobbies, I build machine learning and data-centric software that makes life easier.

💡Timothy Dobbins💡

Shaw Industries

Generative AI/Large Language Models/Image Generation Models
•	Led a team of data scientists and engineers to build an LLM chat application using RAG (retrieval augmented generation) and agent architecture
•	Integrated multiple models through Azure OpenAI and Databricks model serving
•	Tool released across our commercial brand for productivity gains, knowledge retrieval, product design, creation of marketing materials, and general AI use case testing/validation

Machine Learning/Predictive Modeling
•	Led a team of data scientists, engineers, analysts and stakeholders to build a machine learning model that provides product samples placement recommendations at the individual customer level based on the probability of a positive ROI
•	Model uses SMOTE to handle imbalanced data and decision tree classifiers for the predictions
•	Results are integrated directly with Salesforce and used by Shaw’s entire sales force to make placement decisions
•	Model estimated to save company $4-5M annually

Generative AI/Large Language Models 
•	Led a team of data scientists, engineers, business analysts and stakeholders to build a fully functioning prototype that extracts order information from free form text in customer emails and places the orders automatically through our internal order application’s API
•	Solution includes Microsoft Power Automate for email detection and ingestion and large language models (GPT-4) for order information extraction
•	Secured funding and buy in for full scale implementation. Currently leading the team through development

Cloud Platform Migration and Administration
•	Pioneered and oversaw the transition of the company’s data science tools to a standardized stack primarily leveraging cloud technologies. Tool chain includes: Azure, Databricks, Git/Github, Posit Connect
•	Oversee team that handles platform administration as well as all user management, onboarding, training and issue management/troubleshooting

Data Science Manager

Deep Learning/Transfer Learning/Computer Vision
• Built a product similarity engine that identifies closest product alternatives based on the product image and characteristics utilizing both pattern recognition and color matching technology
• Solution was built in Python and includes transfer learning via the VGG16 convolutional neural net (CNN) model, k-means clustering, and image color extraction/distance calculations
• Model results feed directly into our order system platforms and are used by both our sales team and customers to recommend product alternatives for out of stock items

Computer Vision/Machine Learning
• Systematically clustered and classified our resilient products into color categories – providing a way for our sales team to search all products by color
• Sales team converts an additional 1,000 orders per month due to the autoclassification of color categories

Machine Learning/Predictive Modeling 
• Predicted how much sound will travel between floors of a building based on various construction, flooring, and sound wave attributes via a gradient boosted decision tree model in Python
• Model saves the company upwards of $12 million dollars if they were to conduct and measure a sound test for all of our products

Time Series/Forecasting 
• Technical project lead of a team of off shore resources to build new product forecasting models for our luxury vinyl tile product. This involved effectively communicating the business problem, supplying and verifying data, collaborating to design and architect the solution, reviewing and approving modeling methodologies, and going through extensive code reviews for approval and ownership transfer
• Solution involved k-means clustering, dynamic time warping (DTW) and xgboost models

Machine Learning/Predictive Modeling
•	Used a vanilla feed forward neural network to accurately identify which claimants on long-term disability will return to work that had been classified by specialists as highly unlikely to recover (extremely imbalanced data)
•	Model is used to understand future financials and to manage claims

Statistical Analysis/Feature Engineering
•	Calculated durational and occupation-based recovery metrics (average recovery times, probable likelihood of recovering, and relevant percentiles) for each ICD10 code (disease) based on health and claims data. Results systematically computed and stored for reports and for use as model features

NLP/Text Analytics 
•	Created a dashboard for internal auditors that flags expense items that have violated company policy using a combination of NLP methods and rules-based coding

Data Analytics Consultant

Machine Learning/Simulation Modeling
•	Used a combination of predictive GLMs and stochastic simulations in R to forecast insurance claims incurred and compute confidence intervals/probabilities we would be within projection for all our product offerings
•	Final results are used across finance in the calculation of economic capital and 1-in-20 stress impacts

Assumptions Based Modeling
•	Modeled Long Term Care product in SunGard’s Prophet based on a potential external party’s view. Model is used to evaluate reinsurance quotes and better understand financials/product behavior
•	Modeled extreme, adverse scenarios including a U.S. pandemic outbreak and a massive policy lapse     
                                                                                                                 
Statistical Analysis
•	Created a report that groups similar long-term disabilities and determines average recovery times and relevant percentiles. Analysis is used by operational areas to monitor and understand claim activity

Actuarial Analyst

Data Science Manager with 10 years of experience building and leading the design, development and deployment of machine learning and AI-driven solutions at scale– delivering powerful insights, cost savings and revenue growth across multiple business domains.

Strong communicator and thought leader with a proven track record of successfully building and leading cross functional and highly technical teams. Enjoy collaborating with a wide range of business partners.

Deployed projects include generative AI applications that streamline content creation and knowledge retrieval, advanced product recommendation engines that enhance digital customer experience and increase sales, and a multitude of predictive machine learning modeling projects that optimize supply chain efficiency, help understand and plan for customer buying patterns and behavior, and effectively model risk to improve operational decision making.

Educational background in mathematics, analytics/data science and computer science. Conducted research funded by the National Science Foundation on complex random polynomials. Findings published in the Illinois Journal of Mathematics and presented at multiple mathematics and STEM academic symposiums.

taplio

The Battle

💡Timothy Dobbins💡

Katrina Ferrier

Followers

Posts

Engagements

Likes

Ready to grow your LinkedIn presence?