I'm an infrastructure engineer who likes the parts of the job most people quietly avoid: log pipelines, runbooks, the second post-incident review, the cleanup PR no one notices.
My day-to-day is mostly Linux on AWS — keeping fleets of EC2 instances healthy, watching them through ELK and Grafana, and writing the Ansible & AWX automation that holds it all together. I treat observability as a design problem, not a logging problem.
Outside the SRE work, I run a multi-strategy options trading bot for the Indian markets (NSE / MCX). It's the same discipline, different domain — runbooks, kill switches, postmortems, and a healthy respect for the failure modes you didn't anticipate.
I write here because I keep solving the same problems twice. This site is mostly a notebook to my future self, with a side hustle as a portfolio.
Multi-strategy Python options bot, deployed to AWS Mumbai via AWX with a static EIP for SEBI broker whitelisting. Five concurrent scanners, an LLM-backed entry gate, and a session-wide kill switch.
Designed & shipped an Ansible AWX playbook that rotates the apadmin ed25519 key across env-classed AWS RHEL fleets — verify-before-commit, per-host rollback, AWX credential write-back.
Two t2.micro RHEL EC2s, one Ansible playbook to deploy, one to tear down. Logstash + ES + Grafana, dashboards versioned in git and pushed via the Grafana API.
A Linux security reviewer (stdlib-only Python), a Flask + SQLite project tracker, a multi-agent research SDK on Claude, and an AWX dispatcher playbook for EC2 lifecycle. Built to scratch real itches, kept small.
Things I reach for first. Roughly grouped, deliberately incomplete — I'd rather use one tool well than five tools poorly.
I'd rather pick a tool that has been quiet for ten years than one that's exciting this quarter. Boring is a feature; surprises live in the bill and in pager duty.
If I can't see it, I can't fix it. Every system I touch grows a dashboard before it grows an endpoint — including the trading bot.
Cost is a real constraint, not a footnote. The bot runs on ~$30/mo of EC2; the ELK stack tears itself down end-of-day. Most of my favourite designs came from someone forwarding me an AWS bill.
If it's worth deploying, it's worth a one-page document explaining what to do at 3 a.m. when it breaks. Trading systems are the same: a kill switch, a clear way to recover, a written post-incident.