GenAI and Worker's Compensation Claim data science strategy lead

Led a team and projects aimed at explaining, predicting, and prescribing action for customer behavior and feedback. Heavy use of deep learning and NLP. Mix of business-facing products and POCs.

• Managing team of data scientists ranging from 3-6 with some "dotted line" relationships, along with a javascipt developer on certain projects• Team developed novel methods for topic modeling, using pretrained transformers. Models are giving CX group newfound granularity into customer feedback (keras, sklearn)• Leading implementation of a NLP-backed search tool, developed in collaboration with another research team. The search tool is allowing CX and Operations groups to better size feedback (keras, sklearn, Flask)• Developing deep learning models to predict behavior from customer interactions. Including models to predict cancellations and calls; and autoencoders to find anomalous behavior (keras, pytorch, gensim)• Developed internal UIs and RESTful APIs for tagging, labeling, and validating text for NLP models (Flask, HTML)• Developed and taught training for GLMs, Machine Learning, and Deep Learning

Startup side project that briefly went full time. I learned a lot about drones and startups, and got to build some interesting computer vision MVPs. Ultimately, founder life wasn't for me, and I made an exit as the company pivoted away from computer vision into content management software.

• Led modeling projects, while managing and developing a small team of data scientists• Pitched and developed project to identify deviating medical patterns from bills and claim notes. Used Knowledge Representation Learning for medical codes (ICD/CPT), along with LSTMs and GANs (Python: Keras, PyTorch, gensim)• Led model build for new nation-wide Worker’s Comp. initiative to stratify claim severity and assign alternative workflows, using GBM and RF (Python: XGBoost, scikit-learn, pandas; SAS; SQL)• Championed the increased use of text mining at Travelers by teaching others and improving three existing models, adding 10-50% lift in fit metrics (Python: Keras, scikit-learn, gensim, NLTK)• Hosted modeling competition with 52 competitors. Developed submission and leaderboard web services (Python: Flask)

• Led modeling projects, while managing and developing a small team of data scientists• Led model build for new nation-wide Worker’s Comp. initiative to stratify claim severity and assign alternative workflows, using GBM and RF (Python, SAS and SQL)• Championed the increased use of text mining at Travelers by teaching others, and improving existing models using N-Grams, Topic Modeling with LDA, Word2Vec, and LSTMs using claims notes (Python, R, TensorFlow, Keras)

• Led modeling for General Liability initiative, pairing claims with appropriately experienced claim handlers to improve resource allocation, using GLM and GBM (SAS, R and SQL) • Completed 3 month rotation in Worker’s Comp. business team. Gained insight into their needs, and brought new statistical rigor and ideas to their work

• Built models to quantify the impact of new claim initiatives on cost of Personal Insurance claims. (SAS, R and SQL)

• Built models to quantify the impact of new claim initiatives on cost of Personal Insurance claims. (SAS, R and SQL)

Performed research and data analysis with SAS for both internal assessment and external review. Worked with data to compare faculty credentials and subject of courses taught.

Performed assignment grading duties for ST 509 - Statistics for Business Applications.

Worked to uphold a healthy residential community, conducted programming and mediated disputes between residents. Peer elected.

Built and maintained websites using ColdFusion and Microsoft SQL. Designed both user-interface and administrator back-end for publication websites and employee/advertising management systems.

Assisted the professor by answering student questions and administering examinations for PHL 100 - Intro to Philosophy.