K.Boopathi

Full Stack Developer · Infrastructure · Voice AI

Hi! I'm K. Boopathi, a passionate Full Stack Developer with a love for tackling complex problems and staying ahead of the curve in the world of Web Development, DevOps, Generative AI, and beyond.

In this space, I share not just my professional experiences, but the journeys of curiosity and discovery I embark on daily. Think of this as my digital "second brain" where I document everything from debugging adventures to creative AI hacks.

Read the case files Explore my second brain Hire me

Case files

Every one of these started with me staring at a dashboard at an hour I’d rather not admit to, with no idea what was going on. I write them up as detective stories because that’s honestly how they felt a symptom, a trail of confident wrong guesses, and eventually the culprit. Usually it was me.

We declared war on our own database

One innocent-looking command on a live collection of 200 million records. Connections spiked 2,900% and the whole database stopped negotiating.

MongoDB · Self-inflicted The AI that queried a million documents

Our agent called a tool with empty parameters. That became find({}) across a million documents inside a 1GB pod. Trusting the model was the real bug.

Voice agents · 24-hour hunt 233,145 users, one ₹10 biscuit

I scanned a cashback QR code out of curiosity. One query parameter later, I was looking at a quarter of a million people's personal data.

Security · Curiosity I blocked the attacker. It was me.

A thousand requests a second hitting the landing page. I started banning IPs like a digital bouncer and banned our own internal network.

Incident · Humbling A million NXDomain errors

Kubernetes was quietly generating over a million failed DNS lookups. The culprit was five characters of config nobody had ever read.

Kubernetes · DNS Heroku to Kubernetes, the honest version

Every mystery the migration handed us, collected in one place: cold starts, race conditions, probes that lied, and jobs that restarted themselves.

Migration · Series hub

What I go deep on

Real-time voice agents

Twilio media streams, sub-500ms latency budgets, STT and TTS trade-offs, turn-taking, prompt caching and what it actually costs. My most complete original work.

Deepest expertise Doubling a cache hit rate

A 7,200-token prompt, a 1,024-token cache floor, and first-turn hit rates that went from 24% to 62% on outbound calls.

Performance Inside ElevenLabs' agents

I pointed their conversation agent at my own LLM proxy to read the system prompts they send. Here's how it actually works.

Reverse engineering Prompt engineering, mechanically

Not "try this prompt". How tokens, context windows and post-training actually shape what you get back.

From first principles The non-English LLM tax

Hindi costs roughly five times the tokens English does. At 10,000 messages a day that's an $82,000 gap nobody budgeted for.

LLM economics Four years at a startup

Intern to team lead to infrastructure. The senior who told me "I am the documentation", and everything else nobody warns you about.

Career

Second brain

I learn something new almost every day, and for years I forgot most of it just as fast. Notes scattered across three apps, bookmarks I never opened again, concepts I’d understood once and then had to learn all over again.

So I stopped collecting notes and started connecting them. Every idea links to the ideas it sits on top of, the way my head actually stores things by association, not in folders. When I come back to one note the neighbours come back with it, which is what turns revision from an evening into a few minutes. Right now it runs deepest on AI engineering and voice agents, because that’s what I’m in every day.

Open the graph and follow the links, or start from a topic:

Artificial Intelligence Generative AI System Design Databases Books

Let’s connect

I’m an introvert right up until the code breaks then I’m loud, opinionated and completely in. If you’re chasing something strange in production, building voice agents, or you just want to tell me I’m wrong about Kubernetes, I’d genuinely like to hear it. Pick whichever door suits you:

LinkedIn

The fastest way to reach me. Tell me what you're building and I'll reply.

GitHub

The code, the half-finished side projects, and commits with timestamps I'm not proud of.

YouTube

For the things that are easier to show than to write down.

Dev.to

These posts, mirrored, if you'd rather argue in a comment section.

Medium

Same writing, for people who live over there.

Buy me a coffee

Only if something here saved you a bad night. Never expected.