Applied AI & Platform Engineer building the infrastructure that makes LLM agents observable, safe, and auditable — on a foundation of 150+ EKS clusters, 30K+ pods, and a 500+ GPU inference platform. Every revision shipped, none rolled back.
Bengaluru, India · Jun 2026 → present
Lead the design and delivery of an integrated agent observability platform — one place to monitor LLM agents in production and catch behavior that isn't acceptable for enterprise use.
Built detection into the agent harness for PII/PHI exposure and leaked secrets across prompts, tool calls, and responses, so sensitive data is caught before it leaves the agent.
Integrated red teaming to probe agents for unsafe or policy-violating behavior before and after release — turning ad hoc testing into a repeatable safety check.
Unified traces, safety findings, and policy violations into one view so engineering and governance teams can audit agent actions end-to-end and act quickly.
Pune, India · Jul 2024 → Jun 2026
Architected a Kubernetes-based LLM inference platform on Spot-backed GPUs — scheduling, fault-tolerance and retry strategies sustaining ~80% utilization across 500+ GPUs and driving a ~60% compute cost reduction for large-scale batch AI workloads.
Platform enablement for 40+ engineering teams on Amazon EKS: 150+ clusters, ~1,000 nodes, 30K+ running pods. Designed the centralized ArgoCD GitOps platform managing Kubernetes add-ons for 120+ applications across the fleet, with Argo CD Notifications wired across 100+ projects.
Implemented Amazon Managed Prometheus + Grafana observability for 150+ clusters and ran an internal uptime platform covering 2,500+ endpoints — control-plane, ingress, infra and application metrics, with ingestion aggressively trimmed to keep costs low.
Researched and deployed 50+ Kubernetes add-ons (MinIO, Wiz, Kafka, Redis, Dgraph, S3/EFS CSI, DCGM exporter); built application-focused Terraform modules provisioning 80+ AWS resources through a single pipeline; prototyped EMR on EKS with Karpenter for Spark workloads.
Pune, India · Sep 2022 → Jul 2024
Deployed and operated 100+ applications on Amazon ECS with fully automated CI/CD, and delivered 250+ application deployments on EKS via TeamCity, ArgoCD, Helm and GitOps — making deployments repeatable and boring as teams and clusters multiplied.
Built bulk-upgrade automation patching 500+ CI/CD pipelines in parallel, replacing manual, error-prone updates. Led large-scale migrations — EC2 → ECS/EKS and ECS → EKS — untangling legacy assumptions along the way.
Owned 20+ COTS platforms (Sysdig, Dependency Track, ArgoCD) across ECS/EKS; led the Sysdig POC and enterprise rollout across 500+ products in ~1 month using automated rollout tooling.
Pune, India · Nov 2021 → Sep 2022
Owned day-to-day AWS operations and platform lifecycle; built server bootstrapping automation for Windows and Linux; wrote cloud deployment standards and governance guidelines; tuned resource allocation for cost; acted as escalation point for production issues beyond the runbooks.
NVIDIA
GPU and accelerated-computing fundamentals for AI — data center infrastructure, GPU cluster operations, monitoring, and AI workload management.
Anthropic · Credly
Designing and building production-grade applications with Claude — Claude Code, the Agent SDK, the Claude API, and MCP.
platformengineering.org
Platform-as-a-product practice — golden paths, developer experience, and internal developer platform design.
MIT-ADT University, Pune · graduated Aug 2022 · GPA 7.9