Dharshan S Kotian Dharshansk16

Hi, I'm Dharshan.

Backend-focused full stack developer based in Udupi, Karnataka, India. I like systems that scale, retrieval pipelines that don't hallucinate, and code that ships.

Currently interning at Hewlett Packard Enterprise, building a multi-agent RAG system Also at Inspirante Technologies, engineering a hostel management platform. B.E. in Computer Science at NMAMIT.

Lately deep into AI systems — RAG, MCP servers, LangGraph agents, and connecting LLMs to real tools.

Some things I've built

Sophia — AI assistant platform where users train custom personas on their own data. Hybrid RAG architecture combining Neo4j knowledge graphs with vector search to cut hallucination rates and improve retrieval accuracy. (Next.js, LangChain, Azure AI Search, PostgreSQL, Neo4j)

Hackfest '26 Evaluation System — Scalable evaluation interface handling 3000+ participants and 600+ submissions. Built score normalisation logic to reduce bias between strict and lenient judges. Cut evaluation time by 60%+. (Next.js, TypeScript, TailwindCSS, Drizzle, PostgreSQL)

Doc-AI-Agent - Built to address a real enterprise problem: teams spread across Jira, GitHub, Miro, and Confluence with no shared source of truth. Designed a LangGraph agent that classifies documents into semantically related groups, building a centralised knowledge layer across tools. Cascades through vector search → LLM → MCP-powered agent based on confidence, deciding whether a document joins an existing group or creates a new one. (LangGraph, MCP, Python)

Stack

Languages — Python, TypeScript, C++

Backend — Django, FastAPI, Node.js, RabbitMQ

Frontend — Next.js, React, TailwindCSS, ShadCN

Databases — PostgreSQL, Neo4j, Redis, MongoDB

AI & Retrieval — LangChain, LangGraph, fastMCP, ChromaDB, FAISS

Cloud & DevOps — Azure, Docker, Kubernetes, GitHub Actions

Currently into

Building — multi-agent RAG systems with hybrid retrieval (vector + knowledge graph), MCP servers that connect LLMs to real enterprise tools, and stateful agent workflows with LangGraph.

Learning — distributed systems fundamentals (consensus, fault tolerance, message queues), system design patterns at scale, and how the pieces actually fit in production.

Grinding — DSA and algorithms consistently on LeetCode.