Completed two internships with the ads AI team, building and maintaining models to control bids for real time ad auctions

Main project: Implemented and deployed fully automated training service in AWS cloud supporting all our models

Training and evaluation code written in Python using tensorflow

Networked using gRPC microservices, backed by MySQL and DynamoDB

Supported distributed training on CPU or GPU instances using horovod

Fully automated dev and prod deployment using terraform

Fully automated hyperparameter tuning and model selection

Full tensorboard integration for visualization and monitoring of model parameters and performance

Implemented feature pipelines in Scala using Spark to process tens of GB of auction logs per day for model recalibration

Integrated CI-controlled doc generation into team repositories and automatic doc upload into private confluence server

Did not receive full time offer at end of internship due to Covid related headcount decrease

Presented lectures and lead programming labs for classes including compiler design, comparative programming languages, and algorithm design and analysis.

Designed and implemeted a suite of applications to automate protein data analysis, >100 times faster than the previous state-of-the-art on standard benchmarks. Presented at the National Science Foundation TRIPODS conference.

Developed a novel graph matching algorithm for analyzing chemical data. Designed and implemented a graphical interface.