Harshit Luthra

Own platform reliability, observability, incident response, and customer success for a multi-tenant SaaS platform serving enterprise ML workloads across AWS, GCP, Azure, and on-prem. Partner directly with global enterprise customers on onboarding and long-term operational health.

• ·Built a modular Terraform framework that cut client onboarding time by 70% across AWS, Azure, GCP, and on-prem; complemented by AWS/GCP/Azure Marketplace listings that drove a 40% increase in self-service customer acquisition
• ·Migrated the platform's logging stack from Grafana Loki to VictoriaLogs: 94% lower query latency, ~40% smaller storage footprint, 3x ingestion throughput, ~50% lower CPU/RAM; published as a public engineering benchmark
• ·Led a platform-wide monitoring overhaul: bifurcated alerts into P0/P1 severity tiers and migrated critical components to New Relic after a successful PoC, improving signal quality with flat observability spend
• ·Architected a severity-tiered incident management and on-call system integrating Sentry, Grafana, and New Relic with Zenduty and Slack, with team-wise routing across five functional domains, materially reducing MTTR and on-call noise
• ·Drove infrastructure standardization across the fleet (security contacts, K8s labels/annotations, resource conventions) and hardened multi-tenant SaaS via tighter tenant isolation, namespace-scoped RBAC, and resource governance
• ·Escalation point for complex multi-cloud production debugging across EKS/GKE/AKS (Karpenter, EFS/CSI, GPU node scheduling, IAM/IRSA, networking, airgapped artifact registries) while keeping enterprise SLAs intact
• ·Routinely support late-IST and weekend windows for onboarding, cluster upgrades, and live debugging with global enterprise customers (e.g., Riot Games, Zscaler)

Architected and operated AWS-based infrastructure on self-hosted Kubernetes supporting 4M DAU at 1M RPM peak. Built monitoring on RED/USE methodologies with Prometheus, Grafana, Elastic-APM, Pyroscope, Loki, and Robusta.

• ·Reduced annual cloud spend by $200K via self-hosted services, efficient pod/node binpacking, spot-instance utilization, and trimming network flow logs and metrics volume
• ·Designed and maintained highly available multi-broker Apache Kafka clusters and a multi-node ELK Stack on spot instances, balancing data resilience with cost
• ·Hardened cluster security through comprehensive network policies and RBAC configurations, materially reducing the attack surface
• ·Cut deployment times to under one minute using self-hosted GitHub Runners on spot instances integrated with ArgoCD and Devtron
• ·Expanded global reach by rolling out IPv6 across the stack and integrating AWS Global Accelerator to reduce latency for international markets
• ·Achieved 99.99% uptime for stateless workloads on 95% spot instances via time-based scaling and node hibernation during low-traffic windows
• ·Led integration of a full observability stack: centralized monitoring, APM, distributed tracing, logging, profiling, and alerting

Managed core infrastructure (OpenVPN, Heartbeat, Elastalert, Grafana) and integrated developer tooling (Wiki.js, Sentry, JFrog Artifactory). Built Ansible automation for deployments, log rotation, backups, and monitoring agents; tuned CI/CD on Jenkins and AWS CodeDeploy.

• ·Drove infrastructure cost optimization that delivered a 90% reduction in operational expenses while maintaining service quality
• ·Implemented HAProxy load balancing to improve network performance and reduce application response times
• ·Improved developer productivity by integrating open-source observability tools: Elastic-APM, Kafka-manager, Kafka-topics-ui
• ·Designed fault-tolerant data infrastructure on multi-broker Apache Kafka clusters and a distributed ELK Stack running on spot instances

• ·Mentor students in Android and Python Development; review project submissions and provide feedback to improve technical depth and code quality

• ·Performed bug reviews and QA across multiple open-source projects, contributing to improved reliability and security alongside a global contributor community