AI Projects, RAG, Agents & LLMs

An end-to-end AI-powered job application system built entirely in n8n. Uses Apify to scrape fresh job listings, feeds them into an OpenAI AI Agent acting as a Resume Analyst, which rewrites and tailors the resume and cover letter to match each job description with ATS-optimized keywords. The tailored content is injected into a Google Docs template to generate a polished PDF. AnyMail Finder locates the hiring manager's direct email, and a personalized Gmail is sent with the custom resume attached, bypassing the job portal black hole entirely. Dramatically increases application volume while maintaining high personalization at scale.

A medical RAG system built for the NVIDIA Developer Contest using NVIDIA's NV-Embed-QA embedding model and LlamaIndex. Users enter health symptoms and the system retrieves contextually accurate medical insights by embedding documents with NVIDIA embeddings and organizing them via LlamaIndex for fast retrieval. Answers are generated through an intuitive Streamlit interface. Submitted to and recognized by the NVIDIA AI Developer Contest. Makes medical information accessible, precise, and grounded with no hallucinations.

An AI-powered Medical Assistant Chatbot trained on a 600+ page medical book PDF. Uses LangChain, Sentence Transformers, and Pinecone vector database to deliver accurate, context-driven answers. Every response is grounded in the trained document with no hallucinations. Features vector embeddings for precise retrieval, a custom interactive chat UI built with HTML, CSS, and JavaScript, and a Flask backend serving the OpenAI-powered RAG chain. Fully open-sourced on GitHub.

A deployed, production-grade AI assistant integrated into this portfolio. Built with LangChain and OpenAI GPT-3.5-turbo-instruct, it answers questions about my career, projects, and skills using a custom persona-locked prompt with scope guardrails to prevent off-topic responses. Temperature set to 0 for professional consistency. Flask backend with CORS enabled, deployed via Procfile, integrated with the React frontend. This is the chatbot visible on this site right now.

Built with React.js and the OpenAI API, this app uses LLM-powered NLP to predict stock prices and surface estimated returns for investors in real time. Microsoft Copilot Studio handles the conversational layer, letting users query company performance through natural language and receive AI-generated profitability signals without needing financial expertise. The OpenAI model processes market-relevant inputs to deliver structured, actionable outputs through a clean, responsive interface across desktop and mobile.