Building end-to-end AI-powered applications using modern full-stack technologies, including Next.js, React, FastAPI, and Node.js. Specializing in RAG systems, agentic workflows, and LLM integrations to create intelligent, production-ready software solutions.
I'm a AI Full Stack Engineer with a strong focus on building AI-first, production-grade full-stack applications.
I specialize in designing and developing intelligent systems that combine modern web technologies with LLM-powered pipelines.
My experience includes building RAG systems, agentic workflows, and multimodal AI applications using frameworks like LangChain and LangGraph,
along with integrating APIs from OpenAI, Vertex AI, and other LLM providers. I enjoy working across the full stack — from UI development
to backend architecture, vector databases, and cloud deployment.
I am particularly interested in building scalable AI systems, optimizing retrieval pipelines, and deploying real-world applications
that use LLMs beyond simple chat interfaces.
Objective: Build an intelligent system capable of answering questions from PDFs containing text, tables, and images with high retrieval accuracy.
Strategy: Designed a multimodal RAG pipeline using embeddings, vector search, and vision-language models. Integrated CLIP, BLIP, Sentence Transformers, and GPT-based generation.
Objective: Automate end-to-end creation of structured animated videos from a simple text prompt.
Strategy: Built an agentic workflow using LangGraph where multiple AI agents handle script generation, scene planning, voice synthesis, and video assembly using FFmpeg and MoviePy.
Objective: Compare and evaluate multiple retrieval strategies to identify the most effective RAG architecture.
Strategy: Implemented and benchmarked RAG Fusion, HyDE, CRAG, and GraphRAG on a shared dataset with a unified evaluation framework.
Objective: Build production-ready AI applications with full-stack architecture and scalable deployment.
Strategy: Developed full-stack apps using Next.js + FastAPI, PostgreSQL, Docker, and deployed on AWS EC2 with CI/CD pipelines.