# 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 that real people
> use every day. He is available for freelance and contract work.

Site: https://munnasuprathik.com
Contact: hello@munnasuprathik.com

## Who he is

Munna Suprathik is a Generative AI Engineer in Bengaluru who ships production AI
products. He has about a year of production experience with RAG systems, multi-agent
pipelines, and LLM apps. He works across Claude, GPT, and Gemini, and reads the release
notes the day they drop. He holds a B.Tech in Computer Science.

His deeper focus is the harder end of the field: agentic systems and automation. That
means multi-agent orchestration on LangGraph, tool use and function calling, and the
newer agent standards like the Model Context Protocol (MCP) and Google's Agent2Agent
(A2A). On retrieval he goes past plain vector search into reranking, HyDE, and
parent-child chunking, so a model sees the right context and not just the nearest one.
He uses LLM-as-judge to catch hallucinations before users do, and fine-tunes smaller
models with LoRA and PEFT when that beats an API call.

## What he can build for you (services)

- RAG systems and document or knowledge assistants
- AI agents and chatbots
- LLM-powered web apps
- AI integrations and automation into existing products
- Shopify store builds

If you describe what you are building, he can tell you which approach fits and why.

## Projects

- [Tend](https://tendhq.app): A production agentic AI task app. It turns messy brain-dumps into prioritised, scheduled tasks. Built with React, FastAPI, LangChain, the Claude API, and Pinecone. Live with users.
- [Margam AI](https://ai.yourmargam.com): A RAG study-abroad pre-counselling platform. Unbiased, with no referral commissions. Live.
- [Margam](https://yourmargam.com): A study-abroad platform that helps students shortlist universities and navigate the application journey. Margam AI is the AI counselling layer on top of it.
- [Dissect](https://dissecthq.com): An AI SaaS that reverse-engineers viral content patterns, and a second-memory layer.
- ResearchMind: An autonomous four-agent research system on LangGraph (planner, researcher, writer, critic) with reranking, HyDE, and LLM-as-judge. No public link.
- QueryMind: A Mistral 7B fine-tune with LoRA and PEFT. No public link.
- [Quiccle](https://quiccle.com): A curated AI discovery hub. Discover AI in minutes.

## Client work

- [Annore](https://dearannore.com): A gold jewellery Shopify build.
- Esko: Priority 1 security fixes, a jQuery upgrade, and an ASP.NET MVC review.
- [Warangal Margam](https://www.thewarangalmargam.com): A founders and builders community.

## Research (published papers)

- Automated Models in Educational Assessment: A Comprehensive Survey (IEEE, December 2024). A survey of automated grading systems using NLP models like Paragraph Vector, BERT, and a hybrid LDA-BERT approach. https://www.researchgate.net/publication/388779227_Automated_Models_in_Educational_Assessment_A_Comprehensive_Survey
- Enhanced Academic Assessments Using BERT and Top Sampling for Student Performance Optimization (Springer, October 2025). Automates subjective exam grading with BERT for fair, consistent evaluation. https://www.researchgate.net/publication/397778779_Enhanced_Academic_Assessments_Using_BERT_and_Top_Sampling_for_Student_Performance_Optimization

## Notes (writing on building with AI)

Short, honest notes on what breaks in production and how he fixed it.

- [The RAG that retrieved the wrong thing](https://munnasuprathik.com/notes/rag-retrieved-wrong-thing)
- [When I trusted the LLM to grade itself](https://munnasuprathik.com/notes/trusted-llm-grade-itself)
- [The agent that looped forever](https://munnasuprathik.com/notes/agent-looped-forever)
- [Fine-tuning was the wrong answer](https://munnasuprathik.com/notes/fine-tuning-wrong-answer)
- [The prompt that worked in dev and died in prod](https://munnasuprathik.com/notes/prompt-worked-dev-died-prod)
- [Cutting a feature from Tend](https://munnasuprathik.com/notes/cutting-feature-from-tend)

## FAQ

**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 he 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 he 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 at https://linkedin.com/in/munnasuprathik.

**What has he built?**
Live products include Tend (an agentic task app), Margam AI (a RAG study-abroad
platform), Dissect (viral-content analysis), and Quiccle (an AI discovery hub).

**What is his 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.

**Where is he based?**
Bengaluru, India, on IST. He usually replies within a day.

## Contact

- Email: hello@munnasuprathik.com
- LinkedIn: https://linkedin.com/in/munnasuprathik
- X: https://x.com/munnasuprathik7
- GitHub: https://github.com/munnasuprathik