about

Hi, I’m Anurag.

I’m one of those people who can’t quite leave a system alone until I understand what’s actually happening inside it. The thing where you skip the abstraction, peel back a layer, and the world makes a little more sense. That instinct shapes most of my work, and most of what I do when I’m off the clock too.

what i’m into#

• infrastructure & devops — networks, servers, deployment pipelines. The bits that have to stay up while everyone else gets on with their morning.
• self-hosting & the homelab — a quiet little machine under my desk that’s become its own ecosystem. (full tour if you’re curious.)
• networking — routing, protocols, the difference between things working and things mysteriously not working. Packet captures are my crossword puzzles.
• hardware — Raspberry Pis, FPGAs, keyboards, anything I can take apart and learn from.
• AI as a teammate — I use AI heavily for the day-to-day: debugging, log triage, hunting root causes, generating scaffolds, writing tests. Most of my recent side projects (in Go and Rust) were built with Claude Code as the pair sitting next to me.
• linux & macOS — daily drivers. Dotfiles, terminal workflows, the small frictions you only notice once you remove them.

what i’ve built#

A handful of side projects I’m proud of:

• Orva (Go) — self-hosted Function-as-a-Service for homelabs and on-prem. Runs Node.js and Python functions on hardware you own, with sandboxing (nsjail), warm pools, built-in KV, cron, background jobs, and distributed tracing. The serverless experience minus the bill.
• qvm (Rust) — a thin CLI for KVM/libvirt. Docker-style commands for VMs (qvm run, qvm ps), self-contained qcow2 disks, no daemon to babysit.
• SSH-MCP (Go) — an MCP server that lets AI agents SSH into remote servers and actually do things. Useful for the agentic workflows I keep experimenting with.

The rest of my public stuff is on github.com/Harsh-2002.

I’ve also contributed externally — here’s a writeup I did on the OpenObserve blog: Monitoring Caddy, MinIO, NATS and ScyllaDB with OpenObserve.

off the keyboard#

I play Valorant — currently at Diamond 1. If you’re around the same rank and looking for someone to queue with, hit me up.

When I want to switch off the FPS brain, I jump into Forza Horizon — both 4 and 6 (yes, I picked it up day one).

what i’m reading#

A few books I keep going back to — the kind I’d recommend over any blog post on the same topic:

• System Design Interview – An Insider’s Guide by Alex Xu — the standard reference for distributed-systems design. Useful well outside interview prep.
• Ansible for DevOps by Jeff Geerling — practical, current, written by someone who actually runs the stuff he writes about.
• AWS Cookbook by John Culkin & Mike Zazon — recipe-style coverage of the AWS patterns you hit most often. Good for grab-and-go.
• Inference Engineering by Baseten — free, in-depth treatment of serving ML models at scale. Also available as a full-colour print edition from Shroff in India, which is where I read it.

what this blog is#

Notes from the command line. Most posts exist because I spent three hours figuring something out and would rather not figure it out again next year. If a post saves you the same headache, even better.

elsewhere#

• github — Harsh-2002
• x — @anurag_30
• linkedin — anurag-devops
• rss — /index.xml

The DMs are open. Say hi.