Razorpay · Bangalore
Hi, I’m Akhila Nair.
I make production systems calmer.
I work on production systems where downtime is expensive, context is everything, and good automation can change how teams operate. I’m a Senior DevOps Engineer with 6.5 years of experience, currently at Razorpay, building across Kubernetes, reliability, disaster recovery, and AI assisted DevOps.
work
Production systems I’ve worked on.
A few examples from platform and reliability work: readiness, Kubernetes upgrades, and incident review. Open any one for the fuller problem, approach, and outcome.
01 / Go/No-Go at scale Agentic production readiness A readiness flow that gives teams consistent signals before launch instead of another subjective review meeting. Making launch reviews feel less like guesswork. → 02 / Zero-downtime platform work EKS fleet upgrade A safer upgrade rhythm for a large Kubernetes fleet, with checks and sequencing built into the path. Making scary fleet changes feel routine. → 03 / Incident learning faster RCAgpt An AI-assisted incident review tool that helps tired engineers turn scattered signals into a first draft. Making post-incident learning gentler on tired engineers. →
side quests
Little tools I built because I was curious.
This is the part I hope makes people smile a bit: small, local-first projects that came from “wait, why is this annoying?” and turned into useful systems.
01 / Kubernetes release gate kubectl-gonogo
A kubectl plugin for GO/NO-GO release checks and HEALTHY/AT-RISK drift audits on live Kubernetes deployments.
Built in Go for the exact kind of question platform teams ask before a rollout: is this ready, drifting, or risky? 02 / Local-first AI observability aisplaining / AI TraceA zero-dependency Python tool that turns agent traces, tool retries, latency shifts, and cost into plain-English status notes instead of another wall of spans.
Claude Code imports, SDK tracing, OTLP ingest, SQLite storage, and a tiny dashboard when graphs are actually useful. 03 / Your AI week as a story aistoryA weekly AI digest that reads leaderboards, arXiv, and public builder chatter, then turns it into a short story you can actually enjoy reading.
Zero required dependencies, optional LLM narration, scheduled email delivery, and automatic tracing through aisplaining.how I work
I turn operational anxiety into visible systems.
- 01 Make the invisible visible.
Map the actual signals: incidents, logs, ownership, risk, dependency paths, and the places people guess.
- 02 Automate the repeatable parts.
Codify checks, scorecards, upgrade paths, and review flows so the team can move with confidence.
- 03 Keep humans in the hard decisions.
Use AI to compress prep work, not to hide accountability. The machine drafts; the engineer decides.
path here
The short version.
Tresata · Bangalore
Rapid Micro Bio Systems · Lowell
University of Massachusetts Boston
University of Mumbai
tools I reach for
Cloud-native, reliability, and AI operations.
AWSAzureKubernetesDockerHelmKarpenterTerraformGitHub ActionsCI/CDPythonBashPrometheusVictoria MetricsGrafanaCoralogixOpenSearchNew RelicSLO/SLA designLLMsAgentic automationClaude SkillsSubagent orchestrationVector databasesAIOpsMulti-AZ DRRegional DRProduction readinessPlatform scalabilityIncident management
writing
Where DevOps Ends and AI Begins.
Public notes on incidents, agents, dashboards, and the messy human edge of automation.
Part 1 The Night the AI Knew Before I Did The moment AI moved from automation idea to on-call partner. Part 2 You Don't Have to Understand Everything. You Have to Understand Enough. A working model for deciding where AI can act and where humans review. Part 3 Your Agent Can Look Perfectly Healthy and Be Completely Broken AIOps, health signals, and the difference between green dashboards and real readiness. RCAgpt RCAgpt: Helping Incidents Become Stories Teams Can Learn From A public case narrative for LLM-assisted incident review.
if you’re curious
A little more about me.
I’m Akhila. I like the part of engineering where production reality meets new ideas. At work, that usually means Kubernetes, reliability, disaster recovery, observability, incident review, and figuring out where AI can actually help instead of just sounding impressive.
My strongest work happens when a problem is messy and operational: a release needs a clear gate, a cluster upgrade needs a safer path, an incident needs context, or a team needs tooling that explains what changed. I enjoy turning that ambiguity into systems people can trust and use.
I keep up with the latest AI by building with it, not just reading about it. Some experiments become side projects, some become writing, and some quietly change how I think about production work.
contact
Let’s build systems that behave beautifully under pressure.
Exploring opportunities in Bangalore, Mumbai, and Dubai.