IBM TJ Watson Reseach Lab in Yorktown Heights, NY
My research at IBM focused on leveraging Infrastructure as Code (IaaC) technologies to address the issue of hybrid cloud services management. While many automate the provisioning of network, storage, and compute resources in the cloud, solutions do not exist to automate and orchestrate the deployment of services (e.g. backup, patching, security, monitoring, etc.) in hybrid cloud architectures. I created a Cloud Services Deployment platform to manage Hybrid Cloud Services by automatically generating Orchestration Engine Plugins for vRealize Automation, HashiCorp Terraform, Kubernetes, and OpenStack Heat. I helped develop a data model and architecture for a "Services Registry", which details the deployments for cloud backup, monitoring, storage, and patching services. I deployed this service using Kubernetes and Helm Charts in the IBM Services Platform with Watson. I also published and presented at IEEE NOMS conference on April 27, 2018 (https://ieeexplore.ieee.org/document/8406175/).
I additionally mentored an intern on a machine learning and data science project to identify inactive virtual machines (VMs) in the IBM Research private cloud (called ISIS). We analyzed key features such as important processes running on the VM, network connections to the VM, CPU usage, and user login data to determine if a VM was idle. Using these features, we constructed an ensemble model that achieved 95% accuracy (recall) in identifying inactive VMs. Our system is running monthly in the research department and has contributed to the shut down of hundreds of virtual machines, saving the division $2.1 million each year.
