Santa Clara, California, United States
- Developed a Python tool for AdTechs to measure the cost and latency of TensorFlow and PyTorch models on AWS and GCP. - Automated the deployment of resources, monitored latency and utilization, and estimated cloud compute costs for executing models. - Designed various factorized model architectures and identified parameters contributing to latency and cost, enhancing the efficiency of online advertising with machine learning.
- Delivered a security risk calculation algorithm for User Entity and Behavioral Analytics (UEBA) a month ahead of schedule. - Integrated the algorithm into a pipeline that ingests threat detections using C++ and MapReduce. - Reduced testing time by 95% by developing a functional test in C++ with an extensible test fixture, significantly improving team efficiency.
Investigating how blockchain technology can allow households to be prosumers of solar energy - Wrote smart contracts in Go to create a decentralized exchange on IBM’s Hyperledger Fabric blockchain network to allow households to exchange solar energy with each other directly - Curated unit tests in Go using the counterfeiter library to mock interfaces to test smart contracts - Created a JavaScript application to run tests on the smart contracts once deployed to the network
◦ Automated monthly security patches for Amazon Machine Image builds using Terraform to manage an AWS Codebuild that’s triggered by an AWS Eventbridge Rule, saving a week’s worth of time for engineers every month ◦ Streamlined Continuous Integration/Continuous Deployment (CI/CD) pipeline by adding support for services that use AWS Lambda and by automating dependency detection for Go services