Who is Munna Suprathik?

Munna Suprathik is a Generative AI Engineer based in Bengaluru, India. He builds production RAG systems, multi-agent pipelines, and LLM-powered apps, and works across Claude, GPT, and Gemini.

What does Munna Suprathik do?

He builds RAG systems, AI agents and chatbots, LLM-powered web apps, AI integrations into existing products, and Shopify stores. His deeper focus is agentic systems and automation with LangGraph, MCP, and A2A.

Is Munna Suprathik available for hire?

Yes. He takes on freelance and contract work. The fastest way to reach him is email at hello@munnasuprathik.com or LinkedIn.

What has Munna Suprathik built?

Live products include Tend, an agentic task app; Margam AI, a RAG study-abroad platform; Dissect, a viral-content analysis SaaS; and Quiccle, an AI discovery hub.

What is Munna Suprathik's tech stack?

Python, LangChain, LlamaIndex, LangGraph, vector search, Pinecone, Cohere Rerank, HyDE, parent-child chunking, LLM-as-judge evaluation, function calling, MCP, A2A, and LoRA and PEFT fine-tuning. He builds across Claude, GPT, and Gemini.

Notes · Munna Suprathik

The Prompt That Worked in Dev and Died in Prod

By Munna Suprathik, Generative AI Engineer.

It passed every test I had. All ten of them. I wrote all ten. In hindsight that is not a test set, it is a mirror, and mirrors are very encouraging.

Real users showed up with typos, half sentences, context I never imagined, and two extremely polite people who wrote paragraphs. The prompt that was "robust" fell over on inputs barely different from my own. Prompts are brittle in a way normal code is not, because a single reworded sentence can quietly shift the model into a different mode, and nothing throws.

Code on a screen — Green in dev, quietly wrong the moment a real user typed.

So I built the least glamorous tool in my stack: a spreadsheet of real, messy inputs with the behavior I expect for each. Before any prompt change ships, it runs against that set. The rule that changed everything is simple. A change that wins on the easy cases and loses on the hard ones is not an improvement, no matter how good the diff feels.

A prompt tested only on the inputs you imagined works in the demo and nowhere a user actually lives.