Shubham Haranale

▪ Deployed a production-grade full-stack MERN application to AWS EKS using a GitOps approach with ArgoCD, enabling declarative continuous delivery and automated rollback on failed deployments. ▪ Built a complete DevSecOps pipeline integrating container vulnerability scanning (Trivy), automated testing, and image publishing before every EKS deployment. ▪ Configured Kubernetes manifests, Helm charts, and Nginx Ingress for scalable, zero-downtime rolling updates across namespaces.

▪ Implemented a full DevSecOps pipeline for a containerised Netflix clone — integrating SAST with SonarQube, container scanning with Trivy, and automated deployment to Kubernetes. ▪ Configured pipeline gates to automatically block builds on high/critical CVEs, significantly improving the security posture before production releases.

▪ Architected and provisioned a production-ready 3-tier AWS infrastructure (web, app, database tiers) using Terraform IaC — VPCs, subnets, security groups, EC2, and RDS. ▪ Enforced least-privilege IAM roles, tightly scoped security group rules, and encrypted storage policies via Terraform plan/apply workflows to prevent configuration drift. ▪ Automated full infrastructure teardown and rebuild, cutting environment provisioning time from hours to minutes.

▪ Provisioned a complete observability stack on AWS using Terraform — deploying Prometheus for metrics scraping and Grafana for real-time dashboards and alerting. ▪ Built alerting rules for CPU, memory, error rate, and service availability thresholds; integrated alert routing to reduce mean time to recovery (MTTR). ▪ Codified the entire stack as reusable Terraform modules, enabling consistent one-command deployment across multiple environments.

▪ Built a production-grade MLOps pipeline covering data ingestion, feature engineering, model training, hyperparameter tuning, and experiment tracking with MLflow. ▪ Automated model versioning, evaluation, and deployment workflows — enabling reproducible, auditable ML releases with minimal manual intervention. ▪ Designed the pipeline to be modular and CI/CD-ready, with clear separation between training, validation, and serving stages.

▪ Engineered stateful AI agents using LangGraph for customer support automation — implementing multi-step reasoning, tool use, and conditional routing between agent nodes. ▪ Built cybersecurity-focused LLM agents capable of threat analysis and incident triage, demonstrating applied LLM engineering in a security operations context. ▪ Leveraged prompt engineering and agent orchestration patterns to ensure reliable, structured outputs in both production agent workflows.