Vaibhav Kumar.
I build production AI systems: agents, RAG, LLM pipelines, on top of large-scale data engineering. At Plivo, my agents vet US messaging compliance and classify 2M+ messages a day. Before that: 50M+ daily transactions at Razorpay.
The best version of this page is a conversation: email me, I actually reply.
Experience
Four years shipping systems that stay up: agents, pipelines, and the data infrastructure underneath them.
AI & Data Engineering @ Plivo
- Built a multi-agent compliance vetting system: 4 specialized agents (brand validation, content extraction, campaign generation, decisioning) with tool-calling and structured outputs across Claude, OpenAI, and Azure OpenAI. Took accuracy from 62% → 84% and turnaround from days → 6 hours, automating US messaging compliance vetting end to end.
- Shipped an AI support agent: a 12-step orchestration with 11 tools (SQL retrieval, vision-based document parsing, RAG search, Zendesk and Slack integrations) that auto-resolves customer tickets across 7 categories.
- Built a RAG knowledge service from scratch: FastAPI + LanceDB with 14,492 vectors over docs, runbooks, and past tickets; hybrid filtered vector search with quality-based reranking.
- Engineered a fraud classification pipeline processing 2M+ daily messages with parallel batch scanning of 10M+ historical records, cutting manual audits by 70% and response time by 85%.
Technical Consultant @ Razorpay
- Designed payment analytics infrastructure: star-schema warehouse ETL moving 50M+ daily transactions into Redshift, with reconciliation across MySQL, PostgreSQL, and MongoDB at 99.97% accuracy on 200GB+/day.
- Improved dashboard P95 latency 850ms → 320ms through indexing, caching, and query rewrites.
- Owned integrations for 20+ enterprise clients: REST APIs and webhooks at 99.9% delivery across 2M+ daily API calls.
Live on the Internet
Not slides, not repos: things you can click and try right now.
LIVE
fcstudio.co.in e-commerce, in production
A real store with real customers since 2024, storefront to shipping, run end-to-end. Next.js · MongoDB · Razorpay · Shiprocket.
Visit the store
LIVE
The Great Train Robbery WebGL, zero libraries
A single-file wild-west heist in raw WebGL: no frameworks, no build step, runs offline. Built for the joy of it.
Watch the heist
LIVE
lemon an LLM plays the other side
A pixel-art game about talking someone back into the world: an LLM improvises the character, a 159-action engine keeps it honest. Cloudflare Workers · Groq · Gemini.
Talk to them…and a workshop of smaller things, all running
Each one lives in its own repo and deploys itself. Click through.
knowitall
Self-updating news brief that makes me read every headline before I can browse. Backend: GitHub Actions.
hireme
Job dashboard: scores openings against my profile daily across 40+ sources. Free, keyless.
Board
A splittable whiteboard: panes divide like tmux, each toggles between text and freehand drawing.
BlogVlog
8-bit blog platform: GitHub Pages hosts it, the GitHub API is the CMS. Posts are commits.
Webhook Receiver
webhook.site, minimal: a Worker + Durable Object streams incoming hooks live over WebSocket.
Ideas Notebook, 2018
My college notepad of project ideas, finally getting built: disaster analytics, an Enigma machine, more coming.
2027 Sprint Board
A countdown to 2027 wrapped around a 30-day, 30-build sprint tracker. No place to hide.
tracka·packet
Real traceroutes riding real submarine-cable geometry, with a physics layer that catches geolocation lies at light speed.
doobki
Render tests for a game I haven't built: a man falling into the sea, in pixel art and smooth vector, plus a low-poly WebGL2 study.
Ananta
A wisdom chatbot: the Gita, Homer, Plato, the Stoics. Runs on ₹0, cascading three free-tier LLMs.
anubhav
A realtime music visualizer: mic or tab audio, nine styles from Chladni plates to a shoal that swims to the spectrum. Nothing recorded.
Chidiya Udd
The playground classic, digitized: solo, 2-4 players on one phone, or online rooms.
decay-democracy
A data story: calibrate a democratic-erosion detector on documented backsliders, then point it at India.
VK Arcade
The toy cabinet: a pixel-art LLM game, a music visualizer, horror, a WebGL heist, and more. All single HTML files.
The Glow-Up 📈
Every dev has an origin story. Mine is kept live, for humility.
- 2022Weather App still live, adorably ↗ The first thing I ever put on the internet: one search box, one API call, and unreasonable pride. Preserved in amber.
- 2022Sign Language Translator College capstone gone right: real-time deep learning, and a Project of the Year award.
- 2024fcstudio.co.in live ↗ First thing with actual customers and actual money moving through it.
- 2024+Production AI at Plivo Multi-agent systems vetting compliance for real US carriers. The pipelines grew up.
- 2026The Great Train Robbery live ↗ Came full circle: back to building toys for fun, except now the toy has shaders.
- 2026lemon live ↗ The toys learned to talk back: a game where an LLM improvises the other character, live.
Things I've Built
Production systems and deep-learning projects: agents, RAG, and vision pipelines.
Multi-Agent Compliance Vetting
Four LLM agents with tool-calling and JSON-schema outputs vet messaging campaigns end-to-end, with event-driven SQL concurrency control and A/B prompt evaluation baked in.
RAG Knowledge Service
Vector search microservice: LanceDB, Voyage embeddings, per-source chunking, and reranking driven by human quality ratings. Serves three indexes to production agents.
AI Support Agent
A 12-step agent that reads tickets, routes intent, pulls context from databases and RAG, and drafts resolutions across 7 categories.
Sign Language Translator
Real-time sign-to-text: MediaPipe landmarks over 150K+ samples feeding a bidirectional LSTM. 85% accuracy at 30 FPS, sub-50ms latency.
AI Drawing Analysis (DDSL)
Vision models read architectural fit-out sheets; a deterministic calc engine turns schedules into auditable material quantities, cross-checked against parsed CAD files.
Toolbox
What I reach for, grouped by how I actually use it. Don't know one? Click it and you'll get the ELI5.
AI / LLM
Data Engineering
Languages
Cloud & Tools
Leave a Mark
You made it this far. That deserves a trace. Sign the wall: 140 characters, three marks a day, no account, no tracking. I read every one.
Get In Touch
I'm interested in AI engineering roles at the intersection of agents, data, and reliability. My inbox is open.