Designing & building a suite of Python tools to assist admins and engineers deploy a hardware agnostic Slurm 22 cluster. Clusters must support a myriad of permutations of pre-release GPUs, operating systems, CPUs, drivers, vbios, etc. These internal clusters are commonly used by internal teams such as TensorRT and DLSim.

• Built a pipeline to allow GPU tracing software to run on Kubernetes clusters (originally SLURM)

• Built infrastructure to record GPU/CPU/System performance data of deep learning inference

• Connected a react + elasticsearch powered dashboard to MLPerf and TensorRT team DL performance pipelines to track neural networks performance over different chip/driver combinations

• Created a tool to classify DL networks given topology + determine category of DL network (E.g. NLP, CV, etc)

• General internal tool software quality upgrades

• Create and maintain microservices that communicate with SLURM & other internal tools.

• Prototype dashboards in streamlit/jupyter notebooks and develop them into React Apps

• Developed major components of a Python CLI toolchain to deploy slurm 22 to empower other engineers to test the latest and greatest ML Models on the latest and greatest GPUs.

Assistant Instructor for the Code Fellows 201 Class

Migrate Azure Cloud Service application models to Virtual Machine scalable set models on Microsoft Azure

Built Django apps with RESTful APIs and deployed them to IBM Bluemix and Amazon AWS.