Developing ML models and infrastructure to improve semiconductor manufacturing process.❖ Leading project to automate defect classification of silicon wafers and deploy into production manufacturing (GCP)❖ Developed neural network interpretation method to explain defect classification and estimate defect coverage

My PhD was focused on the design and application of novel neural network architectures to problems in genomics, including:❖ Predicting binding between T-Cell receptors and viral protein sequences using a multimodal convolutional neural network architecture. ❖ Developed a neural network interpretation method based on Markov Chain Monte-Carlo (MCMC) sampling to extract scientific insights from trained models.❖ Generative modeling of cancer gene networks using adversarial graph convolutional neural networks.❖ Exploring generative deep learning models, including generative adversarial networks (GANs) and variational autoencoders (VAEs), for imputation of noisy single-cell RNA sequencing (scRNA-seq) data.

B2B company driving digital and AI transformation. I collaborate with customer teams to deliver data analytics and machine learning systems that integrate with client infrastructure.❖ Built and productionized the data pipeline (sklearn, pandas), testing framework, and front-end (HTML, CSS, Javascript) for TextSense, an end-to-end text analytics platform. • Implemented algorithms (nltk, gensim) and visualizations (Altair, Vega-lite) for sentiment    analysis, topic modeling, time series forecasting, and network analysis of unstructured    text data. Developed novel methods for excitement analysis and aspect sentiment analysis.  • Worked with clients in the insurance, sports, consumer electronics, and food industries to    analyze survey and social media data to derive actionable insights into consumer sentiment.❖ Led model testing, improvement, deployment, and monitoring of a continually learning, human-in-the-loop fraud detection system (AWS, XGBoost, sklearn) for the California Community College Tech Center.  • Built a framework to monitor data processing and results from training and serving    pipelines. Diagnosed and resolved a crucial data leakage problem halting deployment.  • Collaborated with domain experts to engineer features and optimize hyperparameters,    increasing precision and recall by up to 20% Monitored the performance of a continually    learning model and the quality of human-in-the-loop labeling. Responded to drops in    metrics through customer engagement and model improvement.  • Developed model ensembling approach to accommodate shifts in data distribution and    combine neural network models (Keras, Tensorflow) with XGBoost to handle both    structured and unstructured data.

Defense startup aiming to bring data-driven processes to the military.❖ Built data pipeline (sklearn, pandas) to ingest, process, and analyze demographic, biometric, and physical performance data surrounding military firing range and tactics exercises to uncover factors of soldier performance and prescribe changes to soldier training.