Experience

DesktopShipper

• United States
• Transportation, Logistics, Supply Chain and Storage
• 1 - 100 Employee

• Data Engineer

  • Jun 2023 - Present



Self-employed

• Data Scientist - Game Theory and Simulation Design

  • Oct 2021 - Jun 2023

  • Architected end-to-end data pipeline to automate joint player distribution collection, lineup projections, and game theory considerations o Harnessed python, selenium, and beautiful soup to create a page-interactive web scrape of RotoGrinders, NumberFire, and FanGraphs to generate implied variance from various stats the underlie a player’s overall median projections o Conducted ETL processes on initially unstructured text data from the various scraped sources to provide… Show more • Architected end-to-end data pipeline to automate joint player distribution collection, lineup projections, and game theory considerations o Harnessed python, selenium, and beautiful soup to create a page-interactive web scrape of RotoGrinders, NumberFire, and FanGraphs to generate implied variance from various stats the underlie a player’s overall median projections o Conducted ETL processes on initially unstructured text data from the various scraped sources to provide data to a mySQL database with raw player projections and JSON files that can create distributions to price single-game player props o Applied K-nearest neighbors and xgBoost (with hyperparameter tuning) to provide probabilistic weights that underlie a play- by-play simulation made up of each node on the decision tree over the life of each play o Implemented a multi-armed continuous bandit and mixed integer programming to select a small portfolio of lineups from 1-3 million possible choices Show less • Architected end-to-end data pipeline to automate joint player distribution collection, lineup projections, and game theory considerations o Harnessed python, selenium, and beautiful soup to create a page-interactive web scrape of RotoGrinders, NumberFire, and FanGraphs to generate implied variance from various stats the underlie a player’s overall median projections o Conducted ETL processes on initially unstructured text data from the various scraped sources to provide… Show more • Architected end-to-end data pipeline to automate joint player distribution collection, lineup projections, and game theory considerations o Harnessed python, selenium, and beautiful soup to create a page-interactive web scrape of RotoGrinders, NumberFire, and FanGraphs to generate implied variance from various stats the underlie a player’s overall median projections o Conducted ETL processes on initially unstructured text data from the various scraped sources to provide data to a mySQL database with raw player projections and JSON files that can create distributions to price single-game player props o Applied K-nearest neighbors and xgBoost (with hyperparameter tuning) to provide probabilistic weights that underlie a play- by-play simulation made up of each node on the decision tree over the life of each play o Implemented a multi-armed continuous bandit and mixed integer programming to select a small portfolio of lineups from 1-3 million possible choices Show less



Tokio Marine HCC

• United States
• Insurance
• 700 & Above Employee

• Sports Analytics Analyst

  • Dec 2020 - Oct 2021

  • Created an NLP injury history model from Rotowire news updates for players across the major sports o Used NLTK in python to recognize sentence patterns to create correct tuples (>99% accuracy) for body part, side, and injury type for large datasets (Over 330000 lines for NBA alone) o Made use of hashmaps to find the most specific and correct versions of injuries as conflicting news reports arose as stories developed o Prepared standardized part/injury/side… Show more • Created an NLP injury history model from Rotowire news updates for players across the major sports o Used NLTK in python to recognize sentence patterns to create correct tuples (>99% accuracy) for body part, side, and injury type for large datasets (Over 330000 lines for NBA alone) o Made use of hashmaps to find the most specific and correct versions of injuries as conflicting news reports arose as stories developed o Prepared standardized part/injury/side tuples to be loaded into a database served by Azure cloud with start and end dates joined from other licensed sources to be used for future injury models • Ran pricing simulations for contractual bonus and prize risk o Created and sampled from relevant distributions to determine the likelihood of a claim being paid out o Allowed underwriters to tweak assumptions in the simulations and rerun certain scenarios for which they wanted to see more specific pricing Show less • Created an NLP injury history model from Rotowire news updates for players across the major sports o Used NLTK in python to recognize sentence patterns to create correct tuples (>99% accuracy) for body part, side, and injury type for large datasets (Over 330000 lines for NBA alone) o Made use of hashmaps to find the most specific and correct versions of injuries as conflicting news reports arose as stories developed o Prepared standardized part/injury/side… Show more • Created an NLP injury history model from Rotowire news updates for players across the major sports o Used NLTK in python to recognize sentence patterns to create correct tuples (>99% accuracy) for body part, side, and injury type for large datasets (Over 330000 lines for NBA alone) o Made use of hashmaps to find the most specific and correct versions of injuries as conflicting news reports arose as stories developed o Prepared standardized part/injury/side tuples to be loaded into a database served by Azure cloud with start and end dates joined from other licensed sources to be used for future injury models • Ran pricing simulations for contractual bonus and prize risk o Created and sampled from relevant distributions to determine the likelihood of a claim being paid out o Allowed underwriters to tweak assumptions in the simulations and rerun certain scenarios for which they wanted to see more specific pricing Show less



Shaptetech Solutions

• Independant Data Science Consultant-Consumer Behavior

  • Nov 2020 - Jan 2021

  •Conceptualized and implemented strategies to test the incentive structures for Shapetech’s clients o Pulled and aggregated customer data from SQL databases to identify trends in consumer behavior o Architected simulations to assess short- and long-term dollar value add of promotional ideas from marketing departments o Built customer churn and product lifecycle models to compare customer product use to company recommendations o Generated data visualization in… Show more •Conceptualized and implemented strategies to test the incentive structures for Shapetech’s clients o Pulled and aggregated customer data from SQL databases to identify trends in consumer behavior o Architected simulations to assess short- and long-term dollar value add of promotional ideas from marketing departments o Built customer churn and product lifecycle models to compare customer product use to company recommendations o Generated data visualization in ggplot2 and plotly to present product and promotion success metrics to executives Show less •Conceptualized and implemented strategies to test the incentive structures for Shapetech’s clients o Pulled and aggregated customer data from SQL databases to identify trends in consumer behavior o Architected simulations to assess short- and long-term dollar value add of promotional ideas from marketing departments o Built customer churn and product lifecycle models to compare customer product use to company recommendations o Generated data visualization in… Show more •Conceptualized and implemented strategies to test the incentive structures for Shapetech’s clients o Pulled and aggregated customer data from SQL databases to identify trends in consumer behavior o Architected simulations to assess short- and long-term dollar value add of promotional ideas from marketing departments o Built customer churn and product lifecycle models to compare customer product use to company recommendations o Generated data visualization in ggplot2 and plotly to present product and promotion success metrics to executives Show less



ShotTracker Inc.

• United States
• Spectator Sports
• 1 - 100 Employee

• Data Science/Computer Vision Intern

  • May 2019 - Aug 2019

  • Created an OCR computer vision app for reading a basketball scoreboard in C++ running on an Ubuntu system o Optimized the current python app in C++ in to read college basketball scoreboards with 100% accuracy at 30 frames per second o Rebuilt a GUI that helped non-technical users take advantage of homography to calibrate ethernet-based cameras for best data delivery o Used Jira Project management (bitbucket, etc.) and git to efficiently develop software in a team… Show more • Created an OCR computer vision app for reading a basketball scoreboard in C++ running on an Ubuntu system o Optimized the current python app in C++ in to read college basketball scoreboards with 100% accuracy at 30 frames per second o Rebuilt a GUI that helped non-technical users take advantage of homography to calibrate ethernet-based cameras for best data delivery o Used Jira Project management (bitbucket, etc.) and git to efficiently develop software in a team setting •Worked to retrain the make/miss algorithm with 120 hz data o Cleaned training and test sets for the new algorithm by combining time series data from multiple servers and prepared it to be fed into an LSTM neural network o Took responsibility for determining necessary features of a robust training set and executing data collection, cleaning, and management Show less • Created an OCR computer vision app for reading a basketball scoreboard in C++ running on an Ubuntu system o Optimized the current python app in C++ in to read college basketball scoreboards with 100% accuracy at 30 frames per second o Rebuilt a GUI that helped non-technical users take advantage of homography to calibrate ethernet-based cameras for best data delivery o Used Jira Project management (bitbucket, etc.) and git to efficiently develop software in a team… Show more • Created an OCR computer vision app for reading a basketball scoreboard in C++ running on an Ubuntu system o Optimized the current python app in C++ in to read college basketball scoreboards with 100% accuracy at 30 frames per second o Rebuilt a GUI that helped non-technical users take advantage of homography to calibrate ethernet-based cameras for best data delivery o Used Jira Project management (bitbucket, etc.) and git to efficiently develop software in a team setting •Worked to retrain the make/miss algorithm with 120 hz data o Cleaned training and test sets for the new algorithm by combining time series data from multiple servers and prepared it to be fed into an LSTM neural network o Took responsibility for determining necessary features of a robust training set and executing data collection, cleaning, and management Show less