Cloud & DevOps engineer in Pune. I build the boring, reliable middle layer — pipelines that ship, clusters that heal themselves, and infrastructure you can read as code.
Self-taught, then formally trained, now doing it on a live production platform.
Working on AlgoFabric, an NLP-driven financial news and sentiment platform for Nifty 50 stocks running on AWS. I audit the live dev and prod accounts, map what's actually deployed against what the docs claim, and rebuild the architecture diagrams from real resources — 100+ Lambda functions, ECS Fargate services, DynamoDB tables, RDS instances and Step Functions across two accounts.
Six months of structured, lab-heavy training across AWS core services, Linux, Docker, Kubernetes, Terraform and Jenkins. Most of the projects below started as assignments here and kept growing after the course ended.
Preceded by a Diploma in Computer Science from MIT Polytechnic (77.49%).
Every one of these runs. Repos are on GitHub with the full README and pipeline logs.
A Flask service on EKS with liveness and readiness probes, plus a rollback controller: if the new ReplicaSet fails its health gate, a bash watcher rolls the deployment back to the last good revision and reports it back to the pipeline. The cluster hero on this page is a simplified model of it.
failed deploy → healthy state in under 60s, no human
Jenkins pipeline that runs build, unit tests, SonarQube quality gate, Trivy filesystem and image scans, Docker build and push, then deploy — the gate blocks the merge, so security failures never reach an environment.
quality and CVE gates block promotion automatically
A CRUD service taken from a jar to a real workload: multi-stage Docker image, Kubernetes manifests, horizontal pod autoscaling, NGINX ingress and a GitHub Actions pipeline — shipped across three versioned releases with zero-downtime rollouts.
3 releases, HPA scaling on CPU, rolling updates
VPC, subnets, IAM roles, node groups and the cluster itself in reusable modules, with remote state in S3 and state locking in DynamoDB so two engineers can't trample each other's applies.
one terraform apply from empty account to working cluster
Prometheus scraping Node Exporter across instances, Grafana dashboards for CPU, memory, disk and network, and alert rules that fire before a box runs out of room — so you find out from a dashboard, not from a user.
host + app metrics with alerting rules
Forked a gym management app, wrote the Dockerfile, and built a GitHub Actions workflow that builds, tags and deploys to EC2 on every push to main. Also automated the fleet's base setup with Ansible playbooks instead of SSH-ing into each host.
push to main → live in one workflow run
Daily drivers, not a list of things I once read about.
I taught myself this stack the slow way — breaking a cluster, reading the events, fixing it, writing it down. That habit turned into a set of notes and repos I still use at work, and it's why I'd rather show you a pipeline than a certificate.
Right now I'm a Cloud Intern at AlgoAnalytics in Pune, reverse-engineering a live two-account AWS platform into documentation people can actually trust. I'm looking for a full-time DevOps or Cloud engineering role where I own pipelines and infrastructure end to end.
Open to DevOps / Cloud engineer roles. Fastest way to reach me is email.