Runs on-prem, not just the cloud
I build systems that live in a data center or at the edge — reproducible provisioning, containers, and high availability, air-gapped when it has to be.
Site Reliability & Platform Engineer · Open to Junior SRE roles · UPES '27
Rohan Batra
I build and operate systems that stay up — and when they don't, I find out why.
I operate a production system where downtime is a safety defect — designing for failure, monitoring it, and root-causing incidents under pressure. Comfortable across Linux, cloud, containers, and the networking beneath them.
- Always-on
- Production system I operate (UPES-ECS)
- 5
- Incidents root-caused end-to-end
- AWS · Docker
- Cloud & CI/CD, in industry
- HACKIITK '26
- Finalist · IIT Kanpur (top ~0.4%)
Built to keep real systems running
My work maps to how infrastructure actually runs — on-prem and hybrid, operated for uptime, over enterprise networks.
Operated for uptime
Health checks, load testing, incident root-cause analysis, and runbooks. I treat reliability as the job, not an afterthought — and I harden the pipeline (OWASP, SAST/DAST) on the way.
Enterprise networking, hands-on
I built UPES-ECS on Juniper SRX, EX, and Mist — now part of HPE Networking — so I've configured the fabric, not just consumed it.
Happiest close to the infrastructure
I'm a systems & platform engineer (UPES CS '27) happiest close to the infrastructure — Linux, containers, the deploy pipeline, and the unglamorous reliability work that keeps a system up. I like owning things in production: health checks, deploys, monitoring, and the runbook that gets you through the incident.
That started on my own hardware — Linux compiled from source, Proxmox, a home lab I actually run — and now extends from a Software Engineer internship (FastAPI on AWS, security audits, CI/CD) to a resilient, offline-first emergency communication system I designed and operate on enterprise networking gear. Along the way: a HACKIITK finalist project at IIT Kanpur and leading the technical wing of my campus developer community.
The stack I reach for
Weighted toward the systems & reliability layer, with real breadth across cloud, backend, networking, and security.
Linux & Systems
- Linux
- systemd
- Bash
- Nix
- Proxmox
- Compiling from source
Reliability & Observability
- High availability / failover
- Health checks & SLIs
- Load testing
- Incident response & RCA
- Runbooks
Cloud & Containers
- AWS (EC2 · ECS · S3)
- Docker
- GitHub Actions
- CI/CD
- Automation
Networking
- TCP/IP
- DNS
- VLAN / QoS
- Juniper (SRX/EX/Mist)
- Tailscale
Backend
- Python
- FastAPI
- Redis
- SQL
- REST APIs
- C/C++
Security & AI
- OWASP Top 10
- SAST/DAST
- OAuth2 · RBAC
- LangChain
- Ollama
Things I've built and operated
A selection from 74 public repositories — a production emergency system, infrastructure tooling, backends, and security/AI work.
UPES-ECS
Flagship · HPE ShowcaseResilient, offline-first campus emergency communication system on an HPE + Juniper networking backbone — designed, deployed, and operated end-to-end: active/standby HA, health checks, load testing, and incident response. Showcased live to Team HPE.
proxmox-auto-install-assistant
★ 6Scripted, unattended Proxmox provisioning that cut setup time by ~80%; Dockerized so the Debian-only installer runs on Arch and other non-Debian hosts — reproducible infrastructure from a config file.
Gap Hunter
HACKIITK '26 FinalistOffline, AI-powered policy gap analyzer mapping security policies to NIST CSF 2.0 & CIS. LangChain pipeline, Electron desktop, 702-test suite (97.7% pass). Top ~0.4% of 1,300+ teams at IIT Kanpur.
second_brain_database
★ 7Containerized personal-knowledge backend: secure REST APIs with OAuth2 + RBAC, Redis caching, and optimized schemas — Dockerized for consistent CI/CD deployments.
nix-config
Declarative, reproducible machine configuration with Nix flakes + home-manager — infrastructure-as-code for my own systems, rebuildable from scratch.
Karmstrot Builds
Android KernelSU build automation across devices — low-level system configuration with SDLC practices for versioning and testing.
How I debug when it breaks
Building is table stakes. Operating UPES-ECS end to end — where a dropped call is a safety defect, not a bug — taught me to find the real root cause fast. Three from the logbook.
“The dashboard is slow.” It wasn't the network.
- Symptom
- Every dashboard felt laggy; the reflex was to blame the LAN.
- Root cause
- A single-threaded proxy serialized every request, and an uncached status endpoint spawned ~17 subprocess calls per hit on one vCPU.
- Fix
- A 24-slot worker pool, a short-TTL cache, and concurrent collectors — turning a cache-miss from the sum of all probes into the slowest single one.
- Lesson
- Measure before blaming the layer everyone points at.
Every call dropped at ~32 seconds.
- Symptom
- Touch-tone input broke and every call died on a fixed 32-second timer.
- Root cause
- An empty external media address — the server advertised its audio stream on an address unreachable across NAT.
- Fix
- Set the correct LAN media address and reload signaling; input and call stability returned.
- Lesson
- “Up, but calls die on a clock” is a media/NAT signature — not a random failure.
Calls that never hung up.
- Symptom
- Over a hundred calls sat “active” forever, inflating every metric.
- Root cause
- When a caller's hang-up was lost, nothing tore the channel down — media-timeout was off and the menu loops by design. A zero receive-count was the tell.
- Fix
- A media-inactivity timeout so dead channels reap themselves.
- Lesson
- Design for the lost-signal case — “no teardown path” is a slow leak.
Where I've worked
-
Nov 2025 – Jul 2026
Platform Engineering Intern · Techymonks · Bangalore (Remote)
Building Python/FastAPI backend services on AWS (EC2, ECS, S3) and wiring automated security checks into CI/CD. Running application security audits — manual review plus SAST/DAST against the OWASP Top 10 — and contributing to WCAG 2.0 accessibility.
-
Sep 2025 – Present
Technical Head · Google Developer Groups on Campus, UPES
Led and delivered technical workshops for 100+ participants, managed cross-functional teams on production-grade initiatives, and mentored students in cloud, system design, and full-stack development.
-
Sep 2023 – Apr 2025
Linux Support & Community Volunteer · LinuxWale, Dehradun
Onboarded 10+ students to Linux (distro selection, dual-boot, driver/hardware troubleshooting), wrote shell scripts to automate post-install configuration, and mentored peers moving from Windows to Linux as a daily driver.
-
Aug – Sep 2024
Cybersecurity Awareness Trainer · Indian Armed Forces Cantonment Welfare Programs
Delivered digital-safety and social-media-safety sessions across three official welfare programs, adapting technical security concepts for non-technical audiences. Recognized with three Certificates of Appreciation.
Certifications & highlights
HACKIITK '26 Finalist · IIT Kanpur · top ~0.4%
TryHackMe · Top 15%
3× Certificate of Appreciation · Indian Armed Forces welfare programs
Eureka! Junior Semifinalist · E-Cell, IIT Bombay
UPES-ECS × HPE
Showcased to Team HPE, July 2026
UPES-ECS — a resilient, offline-first campus emergency communication system on an HPE + Juniper networking backbone — was presented live to Team HPE. Full call flow, operations dashboard, and offline-first architecture.
Explore the collaboration →The day we presented UPES-ECS
Team HPE visited the UPES campus for a live walkthrough of UPES-ECS on its HPE + Juniper networking backbone.
Photos from the HPE visit to UPES · July 2026
Contact
Looking for a Junior SRE / Platform role.
Open to SRE · Platform · DevOps internships now and new-grad roles from 2027 — based in India, happy to relocate to Bengaluru. If you're hiring for reliability and infrastructure, I'd love to talk.