Kate Reiss Email and Phone Number

Professional Shift Recommender- Contributed to the design, development, and deployment of a recommender system for 3,000+ daily users, utilizing XGBoost, AWS Lambda, and H2O- Enhanced user engagement by 60% through personalized recommendations- Implemented a production monitoring tool to track model metrics and identify feature drift- Optimized user experience by filtering notifications based on predictive analysisAutomated Credential Verification with AWS OCR- Led the end-to-end development of an automated credential verification system within 5 months- Successfully productionized the company's second machine learning service- Saved over 145 hours in the first 3 months with zero false positives- Streamlined the verification process by accurately verifying credentials in under 15 secondsResponsible AI- Delivered an informational talk on responsible AI practices to 50 employees, promoting ethical AI development within the organization- Collaborated with the team of 9 ML engineers to improve code review procedures and advocated for comprehensive unit testing in all production repositories

Completed a 14-week immersive data science and machine learning bootcamp with a focus on supervised learning, unsupervised learning, natural language processing,data engineering, and deep learning in Python. Projects built include:Detecting Bias in Language Models: Clustering Word Embeddings from Wikipedia and Twitter Corpora-Created an unsupervised model using DBSCAN clustering on 200-dimensional GloVe word embeddings using Gensim for over 21 billion tweets and a full Wikipedia dump to compare word semantics between the corpora-Identified words with most semantic difference using cosine distance between word embeddings-Used PCA to reduce 200-dimensional vectors to 100-dimensional vectors-Built an interactive app using Streamlit for users to visualize word embeddings and nearest words in vector spaceClassifying Underwater Dive Photos-Built a convolutional neural network using transfer learning (built on VGG16), Keras, and Tensorflow that classified interesting underwater photos from uninteresting photos with an accuracy of 92%Classifying FDNY Cardiac Arrest Outcomes-Developed a classification model that predicted Fire Department of New York (FDNY) cardiac arrest outcomes with an F1 score of 0.67 using only 911 dispatch data, no medical information-Used logistic regression, XGBoost, Random Forest, Decision Trees, K-Nearest-Neighbors, and Naive Bayes algorithmsPredicting South American Hostel Prices-Regression algorithm that predicted hostel prices with a mean absolute error of $4.18 USD-Machine learning algorithms used: LASSO, Ridge, Linear Regression-Identified country as most important factor, with Brazil associated with a coefficient decrease of 5.08 and Argentina with a coefficient increase of 0.71.

-Assessed, treated, and transported over 500 patients of 911 medical andtrauma emergencies in Los Angeles-Collaborated with doctors, nurses, firefighters, and other medical and publicservice professionals