Kyle

Upgraded training server types resulting in a 36% decrease in training time and 47% hourly cost savings before taking into account decreased time spent on training each model Built new services and GraphQL endpoints on a serverless b2b SaaS backend Built a one-click model training and evaluation service that allowed for automatic and user-triggered updates of our core audience models, freeing R&D and engineering from having to manually train dozens of individual models Built a feature store database and Python library to reduce the time spent preprocessing training data by R&D Built a series of binary and multi-class image classifiers to help sort larger datasets for training more powerful models, and used those models to classify ecommerce image types Leveraged AWS’ CDK to build out our serverless data platform to support both R&D and customer success teams, with applications on the platform built in Python Created a web interface with Flask to enable teams to have graphical access to our data platform, replacing a slow 60,000 row Google sheet Built a long-term MLOps roadmap to plan future projects that will reduce time to deploy new models and enhance data and model observability Helped initiate a culture of documentation by setting up and onboarding users to Notion and building onboarding documentation and runbooks to help users efficiently run complex pipelines

Write tutorials for Python programming at all levels, provide technical feedback for other writers as part of the publishing pipeline.

Built a model evaluation tool that automated a previously manual process, reducing time spent evaluating models from multiple dedicated days to a few hours of running in the background. It was also extensible, allowing the computer vision team to easily develop additional metrics for model evaluation Fixed bugs in and optimized an internal deep learning framework Helped form a team to build out a one-stop platform for the entire data lifecycle from collection through labeling, model development, and model deployment. Designed a larger system for each step of the process Assisted in the management of Affectiva’s EMPath internship program by screening candidates, advising and mentoring interns, and developing plans to grow the program by creating an intern-only conference

Productionized a machine learning model on AWS as a scheduled service that emailed another team when it was time to order new blood draw kits in order to reduce waste Designed and built pipelines to ingest and transform publicly available and internal biological data into an in-house knowledge store. Evaluated and employed NLP solutions for structuring unstructured data. Evaluated tools for a department-wide knowledgebase Built and deployed machine learning models with Python, TensorFlow, and AWS Sagemaker to help answer research and business questions Employed Spotfire to create data visualization templates and administered our company's Spotfire deployment

With others, founded a social analytics company to bring our chief project, Nasdanq (later, DanqExchange), the meme stock market to market. In this role, I pitched investors, ran business operations, and built and designed Nasdanq's analysis engine, which listened to social media and used a proprietary algorithm to track the engagement of memes in realtime. This allowed us to provide a stock market game where users could bet on the popularity trajectory of memes and where marketers and analysts could understand macro-level engagement patterns of any image.

Data Engineering * Built and maintained an efficient ETL pipeline with Perl, Beanstalk, Oracle PL/SQL, Microsoft Transact-SQL, MongoDB, and Solr to process large volumes of data. * Assisted in migration from Oracle databases to Microsoft's SQL Server * Streamlined team testing procedures by setting up and documenting Vagrant instances of development servers * Used MongoDB aggregation pipelines to improve the runtime of a process that would fail after over a week of continually running to finish in a few hours

Perform bug fixes for existing applications (C#, WPF, WCF) Help create new features for existing applications Work with data and stored procedures in SQL Server 2012 (T-SQL) Perform application performance tests Work in AGILE/scrum environment

Designed, executed, and presented experiments that investigated the interaction of attentional processes and the auditory system, especially at the levels of the inner ear and brainstem. Additionally assisted in teaching various classes within the Psychology Department.

Assisted in the execution and presentation of olfactory psychophysics research projects in the lab of David W. Smith.