I’m Ty Andrews, a data scientist specializing in large-scale NLP systems and ML-driven product development.
I architect and deploy production-grade AI systems that solve complex business challenges. My expertise spans developing sophisticated NLP pipelines, designing scalable ML infrastructure, and leading technical teams to deliver robust AI solutions across healthcare, insurance, and consumer technology sectors.
Key Projects:
EMChat: Advanced NLP System for Emergency Medicine
Aug 2023 - Present
Production-grade medical AI system featuring custom RAG architecture, achieving EM physician satisfaction on 5,000+ clinical queries. Implemented robust information security protocols, Azure AD authentication, and scalable infrastructure serving 20+ physicians in beta.
Live Demo: https://emchat.app/signup
Details: EMChat Website
DashGPT: Production-Ready RAG Framework
Oct 2023 - Dec 2023
Award-winning open-source RAG implementation featuring real-time inference, comprehensive analytics, and production-ready deployment architecture. Recognized for technical excellence in the Dash-LangChain competition.
Live Demo: https://dashgpt.live
More Details: GitHub Repo
Fortunato Wheels: Turning Used Car Buying from an Art into Science
Oct 2022 - Jun 2023
This personal project aimed to help myself and friends buy used cars with confidence. I scraped over 3M+ car ads over 9 months, trained a regressor and quantile regressors to be able to predict confidence intervals and deployed the model on AzureML. I Built and deployed a website to interact with the model as well as an email notification system using the model predictions to flag cars for friends and family.
Learn More: GitHub Repo
MetaExtractor: Finding Fossils in the Literature
Dec 2022 - Jun 2023
This project aimed to identify research articles which are relevant to the Neotoma Paleo-ecological Database (Neotoma) using machine learning. Extracted data relevant to Neotoma from recently published articles using NLP techniques for named entity recognition (NER). Develped a web app with Plotly Dash for the data to be reviewed by Neotoma data stewards then submitted to the Neotoma database.
Details: GitHub Repo
Open Sourced NER Model: HuggingFace
June365
Oct 2020 - Sept 2023
June365 is a digital personal trainer with a SmartDeck which contains electric motors and sensors to provide tracking and feedback for users. I developed the patented 3D hand measurement system which measures users hand position in 3D space to within ~5cm. June365 used this to develop real time technique feedback system to help users exercise at home safely and efficiently.
More Info
Connect with me here if you’d like to learn more.