Saravana Periyasamy

* developed analytics for throughput and latency on produced NIMs through NIM Factory * Streamlined factory pipelines to lower latency by removing redundant Docker build steps and implementing seamless image promotion from development to staging and production in GitLab CI. * Proposed a model caching approach leveraging an NFS-based file system (RWX) in Kubernetes. * Optimized GPU usage by bin-packing workloads that did not require stringent performance guarantees, introduced NUMA-aware scheduling to handle performance-critical GPU workloads efficiently * Enabled the factory pipeline to create multi-architecture containers for ARM. * Set up Infiband to build and test Deepseek R1 for peak performance * Recommended a design to securely process confidential models in the factory using confidential computing.

Served as an SRE for the MagLev platform team powering autonomous vehicle services. * I optimized Jenkins by prioritizing change requests in Gerrit’s monorepo based on submission dates and set up monitoring to detect deadlocks and large job queues. * Managed the Bazel build system * Introduced an Incident Command process, and rolled out Netflix’s dispatch tool to streamline incident management via Slack—promoting its adoption across the organization. * Created a centralized platform engineering portal using Backstage, providing a single pane of glass for MagLev services in AV Infra. * Implemented ArgoCD to enable continuous deployment of services across multiple Kubernetes clusters, leveraging Terraform for infrastructure as code. * Developed an LLM based RAG (Retrieval-Augmented Generation) slack bot to answer support questions based on our documentation and previous slack threads.

Manage & monitor the cloud infrastructure of Tanzu Mission Control Build tools to help Engineering teams to use the Infrastructure in a reliable way Manage Incidents & conduct Post Incident Review Custom k8s controllers for infrastructure automation Supported the highly fault-tolerant system with cell-architecture Managed cloud cost using ARM instances & spot instances Worked on replicating the tanzu extension container image across regions for reliability & low latency for customers Mentor junior team members https://cloud.vmware.com/tanzu-mission-control

Running multi-tenant Kubernetes clusters in the datacenter & cloud (gcp). Setup load balancers in the cloud using terraform. Wrote custom rpms for kubernetes etcd, master & worker using gradle plugins. Setup spinnaker pipelines to create & deploy immutable kubernetes images in the cloud. Enabled Kubernetes to deploy apps to multi-az in the cloud. Setup cloud autoscaler to scale up/down our worker nodes in the cloud. Wrote ansible playbooks to deploy the immutable kubernetes images in the datacenter. Setup persistence in the Datacenter kubernetes cluster using rook & then moved it to NFS. Setup Prometheus to gather kubernetes infra metrics & as well as app metrics. Wrote Kubernetes custom namespace controller to automate on boarding. Wrote Kubernetes custom F5 controller to automate the load balancer for ingress in datacenter. Wrote Kubernetes ingress admission controller to avoid duplicate & invalid ingress hosts in a multi-tenant setup. Setup MacVlan cni plugin with DHCP to get routable IPs for pods. Contributed a fix to the DHCP plugin - https://github.com/containernetworking/plugins/pull/255 Help app teams to design cloud-native apps & host in the Kubernetes cluster. Help the app team troubleshoot network latency. Trained the team members in the US & India on Kubernetes

Running the cloud platform for Target development teams. We use drone, consul, vault, terraform & spinnaker to build a immutable deployment pattern, self serving cloud platform for all the digital dev teams. Wrote RPMs for sensu monitoring services, and created spinnaker pipelines to deploy immutable images in the AWS cloud.

Migrated the AWS cloud deployments from convoluted chef to modern salt stack, and made it simple with single click deployments Moved the app session to Elasticache to make it stateless Migrated mysql server to AWS RDS

• Migrated the QA environment from citrix to AWS, allowed us to stand up a testing environment when needed and destroy it after use. • Built a VPC, established the site-to-site VPN connection between Best Buy DataCenter & AWS. • Automated the local dev environment using Vagrant/chef-solo. • Automated the cloud deployments using chef, python (boto & fabric) and AWS Cloud Formation Templates. • Developed the High availability Cloud Architecture for the system. • Created Best Practices for DevOps role. • Reviewed the developer’s work and setup the DevOps process for automating the delivery pipeline. • Trained the team members on Vagrant, chef & AWS. • Architected the blue-green deployment on Citrix Environment.

• Worked closely with the Business Analysts, Project and Product Managers to understand the requirements and interpret in the form of user stories. • Lead the team of two .NET developers at onsite and ten .NET/Flex developers at offshore. • Estimated the user story points. • Acted as the Scrum Master for the project team. • Designed and developed user stories using .NET, Java, AWS, RESTful API and Flex. • Automated the build process using Nexus/Maven, Ant and Bamboo. • Explored Chef, a DevOps tool to automate the deployment to cloud.

• Worked closely with the Business Analysts of three applications and managed the monthly delivery of these systems. • Lead the team of three to five .NET/VC++ developers at offshore. • Designed and developed user requirements using .NET and VC++. • Automated the build process using Cruise Control.NET. • Estimated, Planned, Executed and Monitored the Projects using MS Project. • Improved the performance by suggesting and using the Red Gate’s ANTS Performance Profiler & Memory Profiler. • Used Install Shield to package the Spectrum application as windows installable to run on both Windows XP and Windows 7 (32bit/64bit).

• Worked closely with the Energy Marketing group to collect requirements and interpret in the form of use case documents. • Monitored and Maintained SQL DTS packages. • Lead the team of two .NET developers at offshore. • Designed and developed user cases using .NET. • Estimated, Planned, Executed and Monitored the Projects using MS Project. • Deployed builds to production IIS servers and managed the performance of the website. • Technical support for the entire production website involving Application/DB/IIS Servers.

• Responsible for analyzing the requirements from Onsite • Prepared UML class diagram and sequential diagrams using Visio as part of the design. • Prepared the technical architecture documentation and reviewed it with the Cognizant Architects. • Developed many web forms using the 3-tier architecture. • Lead the team of five .NET developers. • Travelled to onsite location (London, United Kingdom), to setup and deploy the build to production environment and setting up the offshore support process.

Out of my own curiosity on AI/deep learning, did this course and had fun! Heres my sample projects - https://github.com/saravanakumar-periyasamy/deep-learning