Site Reliability Engineer · Kubernetes · Cloud native · Open source
I keep production Kubernetes fleets healthy and fix the open source tools they run on.
About · Experience · Open source · Talks and mentoring · Projects · Writing · Certifications · Stack
I'm an SRE based in Pune, India. At Obmondo I ran a GitOps fleet of 50+ production Kubernetes clusters across AWS, Azure, Hetzner and bare metal, and owned the storage underneath it. At the same time I was a maintainer of KubeStellar, a CNCF Sandbox project, where I also mentored LFX contributors. In 2026 I completed Google Summer of Code with the JBoss Community, contributing to jws-diag, a diagnostic CLI for Red Hat's JBoss Web Server.
- 🔧 Now: contributing upstream to Kubernetes SIG projects (Kueue, kind, Kubebuilder), Hive and jws-diag
- ✍️ Writing: debugging stories from Kubernetes and Prometheus on dev.to
- 🤝 Open to: SRE, Platform and Infrastructure roles. Remote in any time zone, or on site in Pune or Bangalore.
| 🧭 Quick facts | |
|---|---|
| Roles I'm targeting | Site Reliability Engineer · Platform Engineer · Infrastructure Engineer |
| Experience | SRE at Obmondo · KubeStellar maintainer and LFX mentor · CNCF LFX intern · GSoC 2026 |
| Production scale | 50+ Kubernetes clusters · 150+ nodes · 8+ enterprise tenants · 524.8 TB of ZFS storage |
| Open source |   |
| Certifications | CKA · KCNA · AWS Solutions Architect Associate · RHCSA |
| Speaking | KCD Delhi 2026 · Cloud Native Pune |
| Education | B.E. Computer Engineering, PDEA College of Engineering, Pune · 2026 · CGPA 8.46 |
| Location | Pune, India (IST, UTC+5:30) |
Site Reliability Engineer · Obmondo · remote · 2025 to 2026
- Ran a multi-cloud GitOps fleet of 50+ production Kubernetes clusters on AWS, Azure, Hetzner and bare metal for 8+ enterprise tenants, shipping 30+ Helm applications through ArgoCD and Puppet across 150+ nodes.
- Owned the storage stack: Ceph (25+ runbooks, OSD recovery in rescue mode, live Docker to containerd migration), ZFS RAID-Z3 pools holding 524.8 TB on 48-disk configs, and Harbor (database moved to CloudNativePG, garbage collection freeing 194 GB+ per cycle).
- Led the RCA for a platform outage caused by two failures at once: Cilium config drift (a missing
--devicesflag) and a ZFS LocalPV DaemonSet scheduling mistake that evicted a Redis StatefulSet. Service was back in 3 hours through ArgoCD reconciliation and node label fixes.
The counts below are live. Click any number to see the pull requests behind it.
| Project | My role | Merged PRs | What I worked on |
|---|---|---|---|
| KubeStellar CNCF Sandbox · multi-cluster Kubernetes |
Maintainer (UI approver, 2025 to 2026) and LFX mentor |  |
Multi-cluster topology views, in-browser pod logs and exec, Helm chart deployment tab. Reviewed 50+ contributor PRs and onboarded 6+ new contributors. |
| jws-diag Red Hat JBoss Web Server (Tomcat) |
GSoC 2026 contributor |  |
summary, config, logs, diff and instances commands, a server.xml parser that applies Tomcat defaults, TLS parsing with credential redaction, and an 80% test coverage gate. |
| Kueue Kubernetes SIG Scheduling |
Contributor |  |
Fixed silently dropped errors in SparkApplication volumes. Tests for the kueueviz backend. |
| Apache Tomcat | Contributor |  |
Unit tests for six valves, including AccessLogValve, PersistentValve and SemaphoreValve. |
| Hive AI agent orchestration for OSS maintainers |
Contributor |  |
Self-hosted Kubernetes fixes: RBAC steps, storage and probe docs, image pull policy. |
| OpenEverest CNCF Sandbox · databases on Kubernetes |
Contributor |  |
Event replay buffer so reconnecting subscribers don't lose events. |
| Apache SkyWalking BanyanDB | Contributor |  |
Dynamic reload of TLS certificates and keys. |
- fix: name xt_mark in the missing netfilter module remediation · hivecommons/hive · merged 30 Sep 2026
- docs: list the InferencePool CRDs as a standalone overlay prerequisite · hivecommons/hive · merged 30 Sep 2026
- fix: pin the CI runner dind sidecar image by digest · hivecommons/hive · merged 30 Sep 2026
- fix: load xt_mark in the node-prep DaemonSet · hivecommons/hive · merged 30 Sep 2026
- docs: check the vllm rollout in hive-inference in manual-provisioning.md · hivecommons/hive · merged 30 Sep 2026
- docs: fix HIVE_VLLM_ENDPOINT comments in the standalone overlay · hivecommons/hive · merged 30 Sep 2026
- kueueviz: add unit tests for cluster queue and cohort handlers · kubernetes-sigs/kueue · opened 28 Sep 2026
- kueueviz: return clusterQueueName and preemption in workload details · kubernetes-sigs/kueue · opened 28 Sep 2026
- kueueviz: add unit tests for local queue, namespace and flavor handlers · kubernetes-sigs/kueue · opened 28 Sep 2026
- fix(tls): find the keystore Tomcat actually uses · web-servers/jws-diag · opened 24 Sep 2026
- FIX: backport duplicate y-tick position fix from waterfall() to waterfall_legacy() · shap/shap · opened 4 May 2026
Selected pull requests, grouped by the kind of work (click to expand)
Reliability fixes
- openeverest#2591: replay buffer so subscribers that reconnect don't lose events
- kueue#13548: stop silently dropping errors in SparkApplication
addVolumes - skywalking-banyandb#642: reload TLS certificates and keys without a restart
- hive#9334: always pull the floating image in the self-hosted Deployment
- jws-diag#65: read listening sockets from
/procinstead of binding ports - jws-diag#67: check the Tomcat process user, not the user running the tool
Features
- jws-diag commands:
summary,config,logs,diff,instances - jws-diag#9 and #10: server.xml parser with Tomcat defaults, TLS parsing and credential redaction
- jws-diag#36: one exit code contract across every command
- kubestellar/ui#357: WDS and WECS topology views
- kubestellar/ui#426 and #434: deploy workloads from Helm charts, popular charts section
- kubestellar/ui#761 and #893: object summary tab, pod logs and exec
Tests and quality
- jws-diag#27: JaCoCo reporting with an 80% instruction coverage floor
- tomcat#965, #976, #979: tests for JsonErrorReportValve, FilterValve, ProxyErrorReportValve, SemaphoreValve, AccessLogValve and PersistentValve
- wildfly-elytron#2422: tests for OidcClientUriBuilder
- kueue#13488 and #13550: kueueviz handler tests, a fixed sleep replaced with a stability poll
Community
- cncf/mentoring#1464 and #1472: LFX mentorship projects for KubeStellar UI (Developer Relations, Model Context Protocol)
- KCD Delhi 2026 (21 Feb 2026): AI-Assisted Kubernetes Policy Management with Kyverno, with Akshay Kumar. Agenda
- Cloud Native Pune (CNCG x Docker): KubeStellar for multi-cluster and edge workload placement
- LFX Mentorship, CNCF: mentor on 4 KubeStellar UI projects in 2025, including the Developer Relations and Community Growth track
- GeeksforGeeks Campus Mantri: ran coding workshops and contests at my college
|
Open source observability platform for Kubernetes with an AI-assisted incident response engine.
Go · Python · CrewAI · Kubernetes · Prometheus · PostgreSQL · SQLite |
- Deleting the corrupted chunk file made it worse · 26 Sep 2026
- Kubelet watches inodes. Just not until it's an emergency. · 17 Sep 2026
 CKA Certified Kubernetes Administrator Verify on Credly |
 KCNA Kubernetes and Cloud Native Associate Verify on Credly |
 AWS SAA Solutions Architect Associate SAA-C03 |
 RHCSA Red Hat Certified System Administrator EX200 |
| Area | Tools |
|---|---|
| Kubernetes | Kubernetes, Cluster API, Helm, ArgoCD, Cilium, Kyverno, Velero |
| Storage and data | Rook-Ceph, ZFS, Harbor, CloudNativePG, PostgreSQL, Redis, MongoDB, MySQL, OpenSearch |
| Observability | Prometheus, Grafana, Alertmanager, incident response and RCA |
| Cloud and IaC | AWS, Azure, GCP, Hetzner, bare metal, Terraform, Puppet |
| CI/CD | GitHub Actions, Jenkins, SonarQube, Docker |
| Languages | Go, Java, Python, TypeScript, JavaScript, C/C++, Bash |
Hiring for SRE or platform work? Email onkarwork2234@gmail.com or message me on LinkedIn.
The merged PRs, PRs in review, posts and contribution graph refresh every day through a GitHub Actions workflow in this repo.