Hai Yu Email & Phone Number

Dallas, Texas, Us

NLP• Developed text annotation system scalable to terabyte data with NLTK, Spark, spaCy, Stanza• Improved occasion extraction model in which consumer use or associate with a product using sequence modeling in Hugging Face with 20% accuracy increase• Researched and implemented product attributes extraction, attributes hierarchy creation and attributes based sentiment analysis algorithm in Spark• Built word subjectivity and opinion strength predictive model with TensorFlow and ConceptNet• Completed consumer segmentation and brand affinity model building using PCA, K-means, OneVsRest classifier in Spark MLlib• Created global e-commerce domain word vector using Gensim to improve downstream tasks• Collected demographics (gender & age) datasets and built deep learning classifier using GloVe, biLSTM and RoBERTa, achieving 15% accuracy improvement• Implemented ETL pipeline to managing complex and frequently changing ETL needs with SparkComputer Vision• Led and established image mining workflow with image collection, image labelling, image preprocessing, image augmentation, image classification (EfficientNet) and image model visualization• Designed dominant color detection algorithm for fashion domain using OpenCV, Image Segmentation and Face Recognization and Clustering Algorithm• Built image similarity query model using Variational Autoencoder

Dallas, Texas, Us

• Built vision-language multi-modal using BERT and VGG16 for trash cleaning and products categorization to make manual work drop from days to hours• Innovated brand clustering algorithm with customized vectorization, hierarchical clustering and locality-sensitive hashing to reduce brand cleaning time by half• Designed domain specific spell checker and canonicalization algorithm with n-gram and fastText• Built brand detection model from product name using transfer learning in Keras• Researched and developed consumer psychometric features extraction and personality analysis method• Topic modeling using LDA, created interactive topic model visualization (LDAvis), and implemented topics hierarchy creation to gain insights from consumer reviews

Dallas, Texas, Us

• Built occasion multilabel classification model to expand occasion extraction capability using scikit-learn• Researched products deduplication, entity resolution, fuzzy matching using machine learning (Dedupe)• Developed Twitter and Facebook collection system with multithreading in python• Audited and crawled hundreds of popular websites using XPath and regular expression• Implemented end-to-end test for angular web applications with Cucumber JS, Selenium Web Driver

Ames, Iowa, Us

• Developed an Android application to mitigate loneliness among seniors by combining Brainwear Wireless EEG Technology and Android; Designed and implemented remote control of devices using Java.• Maintained smart devices and Java programs in Smart Home Lab using OSGi Java framework.• Delivered technical demonstrations and presentations to students, professors, or other visitors.

Ames, Iowa, Us

• Courses: Data Structures in Java, Database systems, Design Patterns, Microsoft Excel and Access 2013.• Designed Java homework and exams and implemented JUnit programs to grade students’ projects. • Tutor or mentor students.• Obtain materials needed for classes, including texts and other materials.• Prepare presentations for lectures and deliver lectures • Help professors and teachers develop course plans• Lead discussion sections.

Des Moines, Iowa, Us

• Resolved record lock issues to control concurrent data access and ensured the integrity of data.• Implemented small-scale programs for Unit tests and implemented automated testing using Fitnesse.• Maintained existing programs and database, and worked on Agile release process in the Scrum team.

Master’S Degree, Computer Science

Bachelor’S Degree, Software Engineering