Cloud & DevOps Engineer

Built and operated a single-node Red Hat OpenShift cluster for internal services and development workloads on a bare-metal RHEL host using KVM, HAProxy, and dnsmasq. The platform supported 15 to 20 microservices running across 40 to 60 pods.

Designed an autonomous self-healing platform that correlated Prometheus metrics, Loki logs, and Kubernetes events. Used large language model-based root-cause analysis and a 57-task Ansible remediation library across 13 domains through Event-Driven Ansible and Ansible Automation Platform, reducing mean time to recovery from 90 minutes to under 2 minutes across more than 20 validated scenarios.

Engineered a Falco and eBPF runtime security subsystem that scored kernel-level events by priority, novelty, and blast radius. Implemented tiered responses including network isolation, forensic capture, and workload elimination, reducing false-positive security alerts by 60%.

Recovered the cluster following a power outage and dnsmasq DNS failure, restoring all 34 OpenShift operators to healthy status within four hours. Also deployed Thanos to provide long-term metric retention beyond the local Prometheus retention window.

Owned the container security track on a three-engineer team. Integrated Trivy image scanning into the build pipeline with policies that blocked critical and high-severity CVEs before deployment, hardened base images, and led remediation efforts that reduced critical and high vulnerabilities by 80%.

Authored Kubernetes manifests and Helm charts for services migrated from legacy virtual machines. Managed application configuration declaratively in Git as the single source of truth and built Prometheus and Grafana dashboards that measured 99.5% availability over three months while confirming a 25% infrastructure-cost reduction.

Investigated and traced a vulnerable OpenSSL package that had been included in a shared base image for months. Removed the vulnerability across all dependent container images.

Built a Python FinOps tool that collected pod-level CPU and memory metrics across eight nodes and 60 to 80 Kubernetes pods. Applied statistical anomaly detection using 14-day rolling windows and generated automated resource right-sizing recommendations.

Identified services over-provisioned by up to three times their actual consumption. Determined that idle development namespaces with full-size resource requests, rather than production workloads, were the primary source of overspend.

Produced recommendations projecting more than 30% compute-cost reduction. Recommendations were partly adopted after the internship.

Engineering degree program focused on Cloud Computing and DevOps.

Preparatory engineering studies in mathematics and physics.