Atharva
Uday Unde
DevOps & Backend Engineer. I build and operate cloud-native infrastructure on AWS and Azure - and the Node.js services that run on top of it.
Five roles, four and a half years, one consistent obsession with uptime.
Assistant Engineering Manager
Leading cloud-native architecture standards across the SaaS suite. IaC-first deployment on EKS, structured reviews, mentorship.
- ▸Established eksctl-based IaC standards, cutting environment provisioning time
- ▸Built an OCR pipeline reducing time required for document processing
- ▸Built Data Import pipelines to import data from different Accounting Softwares
Software Development Engineer L2
Led EC2-to-EKS migration; architected the Node.js microservice suite (paywall, email, file management) on EKS at 100% uptime.
- ▸Provisioned isolated per-env EKS clusters with namespace RBAC and Budget alerts
- ▸Cut lower-priority cluster costs 30–40% via Spot Instances with on-demand fallback
- ▸Resolved latency bottlenecks across two high-traffic services - 20–30% response time reduction
Software Development Engineer III
Cost-optimised AWS via Terraform · led full AWS → Azure AKS migration with zero downtime · Graviton transition.
- ▸40% cloud spend reduction via right-sizing, reserved capacity, Spot usage
- ▸Migrated linux/amd64 → arm64 (Graviton) across production and development
- ▸Built CPaaS/SaaS integrations (VerLoop, WhatsApp Business API) on K8s
Software Development Engineer II
Established Docker-based container pipeline; built high-throughput URL shortener and in-house WhatsApp marketing platform.
- ▸Express + PostgreSQL + MongoDB URL shortener serving hundreds of thousands of redirects
- ▸WhatsApp Business API + AWS SQS marketing platform - saved ~₹70k/quarter vs. third-party
Product Associate
Full-stack POCs for stakeholder demos · contributed to 3 product launches.
The rack.
Paper that backs up the production work.
A few things I've shipped.
K8Lens
Zero-config Kubernetes cluster manager - real-time pod, service, ingress monitoring with health scoring algorithm and multi-arch support.
Meeting Analyzer
Archive Google Meet transcripts, recordings & attendance data beyond retention limits. Free tool for organizing meeting history.
Traefik Dashboard
Reverse proxy management interface - real-time monitoring of services, routes, middleware, and SSL/TLS configuration on Traefik.
Notes from the platform.
Stop Using Cluster-Admin for Everything
If your service account can delete every pod in the cluster, you’re one compromised workload away from a cluster-wide incident.
kubernetes
Stop Storing AWS Keys in GitHub Secrets
Use OpenID Connect to let GitHub Actions assume an AWS IAM role and push to ECR — no long-lived access keys, ever.
aws
The IAM Wildcard That's Waiting to Bite You
A wildcard action on a wildcard resource isn't a policy - it's a promise to get burned later. Here's how to actually scope IAM permissions down.
aws
Container Security: Running as Non-Root
Why most containers run as root, why it matters, and how to fix it in 5 minutes.
kubernetes
"The best infrastructure is the kind you don't have to think about. The second best is the kind you wrote down."
Boring infrastructure
Novelty is the enemy of uptime. I reach for the well-understood thing - Postgres, EKS, Terraform - before the new thing, every time.
Cost is a feature
A 40% bill reduction is the same as a 40% revenue increase, and easier to ship. I treat cost like latency: measure it, set budgets, alert on regressions.
Write things down
Runbooks, ADRs, postmortems. Future-me forgets. Future-team is hired after the fact. The notes are the system.
Small radius of blast
Per-env clusters, namespace RBAC, least-privilege IAM. When something breaks - and it will - it should break in one place, loudly.
Got infra that should be cheaper, faster, or more boring?
I always have time to solve interesting infrastructure problem.