Devatva Rachit Devatva24

About Me

Backend & Systems : Java, Spring Boot, Node.js, Express, Python, REST, gRPC, PostgreSQL, MySQL, Redis
Frontend : React, Next.js, TypeScript, Tailwind CSS
Cloud & DevOps : AWS, Docker, Kubernetes, CI/CD (GitHub Actions, Jenkins), Linux
Data, AI & Observability : SQL, Elasticsearch, machine learning pipelines, Generative AI (LangChain, RAG, LLMs)

LLM Router MCP — MCP server that intelligently routes prompts across Gemini and Groq/Llama based on task intent (planning, codegen, testing, review), minimising token cost without sacrificing quality. Published as an npm package; integrates natively with Claude Desktop, Cursor, and VS Code.
Splitr — Full-stack expense splitting app built with React, TypeScript, and Supabase. Features a minimum-payment settlement algorithm, guest mode with automatic cloud migration on sign-up, UPI deep-link payments, and Row Level Security enforced at the database level. Deployed at splitr-lyart.vercel.app.
Handwritten Equation Solver — Deep learning pipeline combining a CNN (96.97% character accuracy) with an RNN for sequence interpretation to recognise and solve handwritten mathematical equations from images. Served via a Streamlit web app with a live drawing canvas.
Ledger — Personal finance dashboard built with React and FastAPI, backed by MongoDB. Tracks bills, recurring subscriptions with billing cycle normalisation, and tasks with priority levels — with an AI insights layer that flags unused subscriptions and upcoming payment risks.
YouTube RAG Chatbot — AI-powered chatbot that transforms any YouTube video into an interactive Q&A experience using LangChain and Retrieval-Augmented Generation, achieving 25–35% improvement in answer relevance over baseline LLM responses.

Contributing to open source and community-driven projects.
Building and learning backend development with Java, Spring Boot, and RESTful APIs.
Exploring cloud, distributed systems, and applied machine learning.
Open to internships and entry-level roles focused on real-world problem solving.

Based in India · Open to opportunities

_{Thanks for scrolling. If something here resonates with your team, I’d be happy to continue the conversation.}