make $

Google Brain/AI Tensorflow TPU and DTensor Team.• Built next-gen Tensorflow called DTensor, tensorflow.org/guide/dtensor_overview, an open source ML framework for large language models (model parallelism). High performance computing for deep neural networks on TPU and GPUs. DTensor supports tensor sharding through SPMDC++, Python, MLIR, compilers, LLM

FoundationDB Persistence Services Team [Summer 2021]• Built a feature for automatic idempotent transactions for Snowflake's internal record layer above FDB.• Make Snowflake Engineering more robust with idempotent transactions. Impact to many engineers• FDB is a core distributed key-value store with ACID transactionsJava, FoundationDB

Amazon AI Alexa AI Distributed ML, Deep Learning Platform Team [Summer 2020]• Delivered and launched an AutoML [Automated Machine Learning] product built from scratch delivering Alexa impact on scalable, easy-to-use, and framework agnostic Hyperparameter-Optimization for Deep Learning models at Alexa. • Improving Alexa Scientist efficiency and model performance through cloud services and other Amazon tools.• Published white paper to internal Amazon Machine Learning Conference.Used technologies: Python, AWS

Accenture AISystems and Platforms Team [Summer 2019]• Creating a PoC for predictive maintenance with probabilistic ML models backed with HMMs• Predicted failure of sample manual water pumps in Africa within 30 days of true failure• Developed a python package for general-purpose predictive maintenance using Hidden Markov Models • Generate optimized schedule using POMDPs for technicians to repair the pumps• Establish backend using RESTful APIs to expose ML models and set up Docker and databasesUsed Technologies: Python, AWS