Alexander Feng

Machine Learning Engineer skilled in Natural Language Processing with 2 years of experience creating machine learning models and retraining systems and transforming data science prototypes to production-grade solutions. Consistently employs statistical methods and designs to yield real gains from model changes. Effectively researches techniques for novel approaches to problems, develops prototypes to assess viability of approach, and deploys application into production yielding insights to expand customer-consciousness.

Yelp Sentiment Analysis – Built and trained multiple models for sentiment analysis on the Yelp dataset, including a simple RNN, a bi-directional LSTM, and a DAN model. Utilized pretrained word embeddings from GloVe as well as BERT fine-tuning.
Adversarial Trigger Generation for NLP Models – Generated grammatical universal adversarial triggers for Sentiment Analysis that reduce accuracy when prepended to all elements of the dataset and investigated defenses against such triggers.
Hidden Markov Model for POS Tagging – Developed bigram, trigram, and 4-gram Hidden Markov Models for Part-of-Speech tagging using Viterbi decoding