Skip to content

Seven years of making systems predictable.

I'm Shivam Kumar, a platform and SRE engineer. I build scalable, cloud-native platforms that turn complex data into something teams can act on, working across distributed systems, event-driven architecture and container orchestration.

I like roles where the job isn't “keep it up” but “make it resilient”: internal platforms with paved paths, observability people can afford, autoscaling that reads the real workload, and production access that expires on its own.

I design AI agents for complex operational workflows, with a focus on orchestration, safe access and observability.

Where I focus

The problems I want to spend the next few years on, and the ones platform teams keep coming back to.

  • Internal developer platforms

    Self-serve paved paths (templates, GitOps delivery, sane defaults) so teams ship without filing tickets or learning every operator.

  • SLOs and incident learning

    SLOs and error budgets for platform components, blameless reviews, and write-ups that turn one bad night into a runbook.

  • Guardrails built into the platform

    Multi-tenant Kubernetes with RBAC, admission policy and just-in-time access: governance as software, not as a review meeting.

  • Observability you can afford

    Metrics, logs and traces with cardinality and retention under control, so incident-time fidelity survives the budget review.

  • Cost-aware capacity

    Right-sized node pools, Spot and Graviton where they fit, and autoscaling on the signal that actually reflects demand.

  • AI in operations, with guardrails

    Agent-assisted debugging and knowledge capture that stays auditable: journals that become runbooks, not black boxes.

The whole trace

Every company, level and major project on one timeline. Select any span for details.

Career trace23 of 23 spans
Running
2019202120232025

Shivam Kumar

SRE & Platform Engineer

Building scalable, cloud-native platforms that turn complex data into actionable insight: distributed systems, event-driven architecture and container orchestration, run reliably.

Start
Jan 2019
End
Still running
Duration
7y 10mo
Where
Bengaluru
Stack
AWS, Kubernetes, Python, Go, Terraform

Select a span to inspect it. Arrow keys move between spans; right and left expand or collapse them.

Experience

  1. Flexera

    Member of Technical Staff, SRE

    When
    Apr 2024 to now
    Length
    2 yrs 7 mos
    Where
    Bengaluru

    I work on the reliability of Flexera’s multi-region SaaS platform on Kubernetes and AWS, supporting its IT asset management, FinOps and SaaS management products.

    Member of Technical Staff, SRE

    Apr 2024 to now
    • Designed a self-service EKS platform with a central management cluster provisioning isolated workload clusters through Crossplane and GitOps.
    • Designed a declarative Kubernetes access pipeline that separates AWS identity authentication from Kubernetes authorization, with team permission profiles reconciled through GitOps.
    • Designed a TLS and certificate lifecycle for service-to-service traffic across Kubernetes namespaces, covering trust distribution, workload certificate renewal and root rotation.
    • Designed recoverable just-in-time production access with namespace-scoped locks, scheduled revocation and durable notifications.
    • Developed a phased Argo CD capacity plan covering controller sharding, large application resource trees, repository-server sizing and reconciliation pacing, with baseline comparisons before rollout.

  2. MoEngage

    Site Reliability Engineer I to III

    When
    Dec 2020 to Mar 2024
    Length
    3 yrs 4 mos
    Where
    Bengaluru

    Three levels in three years on a high-throughput customer engagement platform, building internal platforms and cutting cost.

    Site Reliability Engineer III

    Feb 2024 to Mar 2024
    • Architected and built an in-house Flink platform on Kubernetes to manage Flink clusters for real-time data processing, with lag-aware autoscaling, auto-healing and performance monitoring.

    Site Reliability Engineer II

    Aug 2022 to Feb 2024
    • Built an in-house Airflow platform so teams could create and manage Airflow clusters on Kubernetes, with monitoring, DAG orchestration and faster rollouts. Cut operational work by 50%.
    • Led the migration of the core monitoring stack from Prometheus to VictoriaMetrics for Kubernetes and EC2, with read, write and ingest tiers that scale independently. Monitoring infrastructure cost dropped 60%.
    • Implemented Karpenter on EKS and moved key SRE components to AWS Graviton to improve cluster performance and reduce cloud spend.
    • Rolled out centralized, real-time logging with Vector, Loki and Grafana, which made debugging containerized applications much faster.
    • Led the Strimzi Kafka operator migration: deep latency troubleshooting, legacy CI pipeline updates and deprecated Kubernetes API fixes.

    Site Reliability Engineer I

    Dec 2020 to Jul 2022
    • Built a Kubernetes-native data warehousing platform on Airflow, AWS Glue, Athena, S3 and PrestoDB, with automation to create and manage PrestoDB clusters and built-in monitoring and alerting.
    • Helped migrate core services serving millions of requests per hour from EC2 to Kubernetes, improving scalability and resilience.
    • Built a GitOps release pipeline with Argo CD, Drone CI and Python automation.
    • Wrote custom validating and mutating admission webhooks in Go to enforce cluster policy and automate resource changes.
    • Managed infrastructure as code with Terraform and built CI/CD pipelines with Jenkins and Drone CI.

  3. Ingram Micro

    Software Engineer, SRE

    When
    Jan 2019 to Nov 2020
    Length
    1 yr 4 mos, plus internship
    Where
    Kolkata

    Where it started: delivery pipelines, cloud migrations and hybrid infrastructure across AWS and Azure.

    Software Engineer, SRE

    Aug 2019 to Nov 2020
    • Designed and implemented multi-stage CI/CD pipelines on AWS and Azure to build, test and deploy microservices.
    • Architected and executed migrations of on-premise infrastructure to the cloud, focusing on performance, security and cost.
    • Built hybrid infrastructure that integrated on-premise data centers with AWS and Azure.
    • Developed an in-house tool that reports on-premise readiness for lift-and-shift to the cloud.

    SRE Intern

    Jan 2019 to Mar 2019
    • First production exposure: pipelines, cloud accounts and the habit of writing things down.

Selected work

What I built, why it mattered, and where to read more.

Multi-cluster platform

Agent-assisted Kubernetes incident response

Designed and built the orchestration path: event normalization, intent-aware routing, remote cluster agents, access grants and observability.

Routes work to an explicit agent capability while cluster grants and Kubernetes RBAC constrain what that agent can do.

  • Kubernetes
  • Python
  • Server-Sent Events
  • RBAC
  • Prometheus
Technical write-up

Flexera, 2025 to now

Just-in-time production access

Issue-driven requests with approval, an atomic per-namespace lock, broker and client pods with absolute deadlines, and scheduled expiry whose cleanup survives notification failures.

Exec access is time-boxed to one namespace and expires without anyone having to act.

  • Kubernetes RBAC
  • Lambda
  • EventBridge Scheduler
  • SQS
  • GitHub Actions
Design write-up

Flexera, 2024 to now

DNS, autoscaling and storage reliability

Traced node-local DNS saturation to ndots search-path amplification and fixed it at the source, sized node pools so one cache never serves 30 workers, and set per-consumer NFS mount policies after diagnosing EFS burst-credit exhaustion.

External lookups go from ten DNS queries to two, and storage behaves the same in every region.

  • CoreDNS
  • NodeLocal DNSCache
  • KEDA
  • Crossplane
  • EFS
DNS postmortemEFS postmortem

MoEngage, SRE II

Airflow platform

-50% ops work

Self-serve creation and management of Airflow clusters on Kubernetes, with monitoring, DAG orchestration and faster rollouts.

Half the operational work it used to take.

  • Airflow
  • Kubernetes
  • Helm

MoEngage, SRE II

Prometheus to VictoriaMetrics

-60% monitoring cost

Moved monitoring for Kubernetes and the EC2 estate to VictoriaMetrics, with read, write and ingest components scaled independently.

More reliable monitoring at a fraction of the cost.

  • VictoriaMetrics
  • Prometheus
  • Grafana

Illustrative upper-bound model with synthetic label counts; no live metrics are queried.

Series upper bound: 15.1M from route × 12 status values × 3 methods × pod × (buckets + sum + count).

MoEngage, SRE II

Karpenter and Graviton

Karpenter for node provisioning on EKS, and key SRE components moved to arm64 Graviton instances.

Better cluster performance and a significantly lower cloud bill.

  • Karpenter
  • EKS
  • Graviton

MoEngage, SRE I

Kubernetes-native data warehouse

Large-scale data processing on Airflow, Glue, Athena, S3 and PrestoDB, with automation to create and manage PrestoDB clusters.

Data teams get warehouse clusters on demand, monitored from day one.

  • Airflow
  • AWS Glue
  • Athena
  • S3
  • PrestoDB

MoEngage, SRE I

Admission webhooks in Go

Custom validating and mutating webhooks to enforce cluster policy and automate resource modifications.

Policy is enforced by the API server, not by code review.

  • Go
  • client-go
  • Kubernetes API
How to build one

Toolbox

Cloud and compute

  • AWS
  • Azure
  • EKS
  • Karpenter
  • Graviton
  • Terraform

Platforms

  • Kubernetes
  • Argo CD
  • KEDA
  • Flink
  • Airflow
  • Strimzi Kafka

Observability

  • VictoriaMetrics
  • Prometheus
  • Grafana
  • Loki
  • Vector
  • eBPF

Data

  • AWS Glue
  • Athena
  • S3
  • PrestoDB

Delivery

  • GitHub Actions
  • Drone CI
  • Jenkins
  • Azure DevOps
  • CloudFormation

Languages

  • Go
  • Python
  • Bash
  • HCL

Certifications

  • Fundamentals of eBPFKernel-level observability and networking
  • AWS Certified AssociateCloud architecture on AWS
  • Microsoft Azure AssociateCloud administration on Azure

Latest from the 3 AM Brew

All issues