• Led team of 5 to build and deploy 2 end-to-end CI/CD ML pipelines for a new client: from ingesting live data from 4+ text sources via Kafka to automating financial risk classification

• Collected requirements from client and organized into discrete pieces of work to delegate

• Performed EDA independently on text, free-text, and numerical input data associated with 2 of these sources and designed transformations like fuzzy matching, keyword search

• Built, optimized, and validated 3 of the 8 decision tree classifiers through several iterations

• Composed weekly reports on the health of the production system, deployed optimizations

• Supported the team’s completion of the rest of the pipeline: integrated with vendors through Kafka, performed EDA, designed and built 50+ table SQL database, designed and built 5 classifiers, designed and built automated testing framework

• Performed technical advising, implemented versioning system and code reviews

• Supported the creation of a question vs. answer classifier for a 2nd client including tokenization, TF-IDF, and decision tree classifier achieving > 96% accuracy

• Replicated 3 papers using CNNs, LSTMs, VAEs, and GANs to predict next frames of video

• Adapted 2 of these models to incorporate natural language inputs

• Automated the localization of contamination using 2D images, 3D point cloud modeling

• Enabled product to scale to new customers by saving inspection of 100s of images per day

• Implemented this system on AWS, fully automating the workflow

• Lead small group problem-solving sessions for 50-person certificate course on data

• Enhanced team’s teaching quality by initiating team meetings and git hub repository