AI Developer & Full-Stack Engineer building intelligent systems at the intersection of LLMs, RAG pipelines and modern web technologies.
I'm a final-year Computer Science student at VIT with a focus on Cloud Computing & Automation. I love building AI-powered applications — from RAG document assistants to computer vision tools — and shipping them as clean, performant web experiences.
My sweet spot is at the boundary of LLM engineering and full-stack development, where I use tools like LangChain, ChromaDB, and React to build products that actually think.
Vellore Institute of Technology (VIT) · 2022 – 2026
Specialisation: Cloud Computing & Automation CGPA: 8.60 / 10.0
End-to-end AI apps using RAG pipelines, LangChain, and Groq LLaMA — from document Q&A assistants to intelligent chatbots with hallucination guardrails.
Responsive, production-grade MERN stack applications with RESTful APIs, JWT authentication, and CI/CD pipelines for seamless cross-browser deployment.
Real-time computer vision systems using OpenCV and MediaPipe — gesture recognition, image stylisation, and AI-powered visual tools with sub-50ms latency.
AWS-certified architecture design, containerised deployments, and environment configuration for scalable, production-ready applications.
Implementing RAGAS evaluation frameworks, hallucination guardrails, structured outputs, and AI quality monitoring systems for reliable LLM responses.
End-to-end RAG application that lets users upload any PDF and query it in natural language, returning grounded answers with source citations, page references, and real-time RAGAS quality metrics (faithfulness, answer relevancy, context precision).
Full-stack NGO website with dynamic pet adoption listings, secure donation portal, and JWT-authenticated admin dashboard. CI/CD pipeline for cross-browser consistency.
Python app controlling system volume via webcam hand gestures — sub-50ms latency, 94% recognition accuracy across varied lighting conditions using finger-distance ratio mapping.
Converts photos to cartoon, sketch, and pencil-colour styles using AnimeGAN2. Integrated Razorpay for in-app payments. Model weight caching cut processing time by ~35%.
The chunking strategy, overlap configuration, and hallucination guardrails that made DocuMind reliable in production.
Read MoreHow I optimised the model loading pipeline at Infosys and the architecture decisions that made the biggest difference.
Read MoreA deep dive into the landmark normalisation and distance ratio mapping that powers the volume controller.
Read MoreI'm actively looking for opportunities where I can apply my AI engineering and full-stack skills. Whether it's a startup needing an LLM app or a team looking for a versatile developer — let's talk.