Cloud resilience
Backward-compatible control-plane changes, staged migrations, regional failover validation, and live-site ownership.
Software Engineer · Hyderabad, India
Software engineer focused on resilient cloud systems.
I’m Harsh, a software engineer at Microsoft working on Azure resiliency. I build cloud control-plane systems, production automation, and carefully operated infrastructure.
01 / Engineering
Engineering focus
My work spans cloud resilience, backend systems, production automation, and operational tooling.
Production automation
AI-assisted incident triage, release validation, test infrastructure, and repeatable workflows that reduce operator toil.
Backend systems
Service APIs, transaction workflows, database correctness, indexing, observability, and performance under concurrency.
02 / Experience
Professional experience
Selected work from my resume. The details are concise here; the full PDF stays current through an automated publishing pipeline.
Download the current resumeMicrosoft
Software Engineer 1 · Azure Resiliency
- Reduced non-production Azure spend by 43%, from $20.7K to $11.8K per measured 30-day period, through policy-enforced governance and cleanup.
- Coordinated backward-compatible API and data-model migrations across 6+ teams, moving 98% of ~18,800 production configuration records with zero throttling.
- Led regional-failover validation across three services, proving uninterrupted reads, 44-second write recovery, and region rejoin after redeployment.
- Built AI-assisted on-call automation that reduced a weekly 4–6 hour workflow to approximately 30 minutes.
KoinX
Backend Intern
- Designed an account-automation engine end to end: MongoDB schema, controller, routing, services, workers, tests, and documentation.
- Implemented transaction-matching and unarchiving microservices, then improved database performance with targeted indexes.
- Introduced MongoDB transactions across critical workflows to strengthen ACID guarantees and data integrity.
Microsoft
Software Engineer Intern
- Automated partner code generation for Azure Migrate Business Case, reducing workload by 30%.
- Built utilities for mocked test generation and Kubernetes operator scaffolding, cutting repetitive implementation effort.
03 / Featured system
Production-style homeserver infrastructure
I designed and operate the architecture, deployment, networking, security boundaries, recovery workflows, and documentation. The applications themselves are upstream open-source projects.
Self-hosted infrastructure · Active deployment
Architecture overview
Public path
01InternetHTTPS clients
02VPS + CaddyTLS edge
03WireGuardNarrow /32 routes
04nftablesPre-DNAT guard
05DockerApproved apps
Recovery path
AAdmin devicePhone or laptop
BTailscaleNative service
CSSH + PiPrivate operations
15running containers
11Compose projects
4access classes
4public upstreams
Network exposure
A Caddy edge proxies only four approved applications over WireGuard. A NAT-aware host guard filters protected ports before Docker DNAT, while databases and support services remain container-internal.
Deployment and drift control
Digest-pinned Compose definitions, rendered host templates, redacted status checks, exact drift detection, and one-stack deployment make the repository the live source of truth.
AI-assisted operations
Pi is available through Tailscale with application authentication and approval gates. The documentation is explicit that approvals reduce mistakes—they are not an OS sandbox.
Incident library
Selected incident reports
Each case study follows symptom → investigation → root cause → resolution → verification.
CASE 03
Caddy over WireGuard
Migrated public delivery to an independently operated edge with narrow routes and application-specific verification.
Read investigation CASE 06Filtering before Docker DNAT
Turned intended access classes into host-enforced boundaries without breaking LAN, Tailscale, or public upstreams.
Read investigation CASE 04One-sided tunnel recovery
Diagnosed healthy services behind a stale peer session, then built a bounded watchdog with age and cooldown safeguards.
Read investigation CASE 07Live source-of-truth cutover
Separated declarations from private state and migrated 11 running projects incrementally with exact drift proof.
Read investigation04 / About
Background
I like understanding software from API contracts and data models down to network paths, deployment state, and recovery behavior. That curiosity shows up at work and in the infrastructure I operate at home.
My default approach is simple: establish the facts, change the smallest useful thing, and verify the behavior that matters.
Education
Manipal Institute of Technology
B.Tech in Information Technology
Competitive programming
Competitive programming
Working set
C# / .NETPythonPowerShellAzureREST APIsCosmos DBMongoDBDockerGitHub ActionsKQLLinuxWireGuardObservabilityAI agents
05 / Contact
Get in touch
For engineering opportunities or a conversation about resilient systems, backend platforms, or infrastructure, email is the best way to reach me.