Developing solutions for Tensorflow on AWS

Worked on Elastic Inference by integrating the product into the tensorflow serving platform, Assisted on the graph partitioning algorithm, created the Continuous integration system, and oversaw the release cadence

Lead the team which created the CI/CD for the tensorflow binaries used in Sagemakee

Worked with the tensorflow team to create a new method of training elastically.

Optimized the training of a large image model to achieve best in time to train.

Develop and maintain SST to assist in Performance Projections for the team, by adding needed features, running simulations, and creating communication models to be run on SST.

Developing performance models for applications which are run on Large scale clusters (Super computers) and project performance to future Large scale clusters.

Investigating Performance issues on both compute and communication.

Rotation 1:

Adding Linux Power management features to an internal system simulator

Rotation 2

Developed a tool to enable users to run studies and perform regression on an internal Network Fabric Simulator

Researched methods to quantify fairness in a Network and also performed studies to quantify fairness within the current network architecture, through which uncovered a fairness issue within the architecture.

Rotation 3

Researched about Big Data, Map Reduce, Spark, and several algorithms that are widely used on these frameworks.

Developed a performance model for an internal framework for Convolutional Neural Networks to better understand the compute and communication needs.