Hey, I am Raghvendra

Currently, I work at Juspay in Bengaluru, on BreezeBuddy.ai, a conversational AI product. I built its chatbot-to-human handoff, which moves a conversation to a live agent with the full context intact, and its guardrail layer, which checks every voice and text reply for prompt injection, toxicity, policy violations and off-topic answers, with rules each tenant can set without a code change.

On the side, I am building Mission HQ, a self-hosted multi-agent system that runs on my own laptop. Its agents share one task board and one markdown vault. They study with me, write my daily note, and draft what I publish here. Nothing leaves the machine without my approval. It is where I test what I read about agents against something that actually runs.

Before Juspay, I interned at Mindtickle on AI and backend systems. I shipped LanguageAndVoiceService, a Go gRPC service that moved voice configuration out of code and into the database, so new voices and languages went live without a deploy, across 25+ languages and 800+ voices. I also built the voice lifecycle pipeline with ElevenLabs and the evaluation pipeline for AI roleplay with Maxim.

Alongside that, I co-founded and built Verly, an AI customer support platform for businesses across web chat, WhatsApp and voice. I built the agent stack in Go: a graph-based RAG pipeline, identity verification, and per-tenant tool access through custom MCP servers, plus the analytics layer that showed us where conversations failed.

I am going deep on how LLM systems behave in production: inference, retrieval, agents, and evals. I test what I read with small labs before I believe it, and I write up what I find in the blog and in short notes.

I studied computer science at IIIT Nagpur, where I spent a lot of my time on competitive programming: Expert on Codeforces, Guardian on LeetCode and 4-star on CodeChef. My full experience and résumé are on the About page.