Embedded Files
AI Document Q&A System (Local RAG)
Overview: A full-stack AI application that allows users to upload documents (PDF, DOCX, TXT) and ask questions through a ChatGPT-style interface using a 100% local LLaMA 3 model (zero API dependency).
The Build: Engineered a complete Retrieval-Augmented Generation (RAG) pipeline. Implemented fast semantic search, optimized document chunking, and source tracking to reduce AI hallucinations and provide accurate answers.
Tech Stack: Python, Streamlit (Frontend), FastAPI (Backend), LLaMA 3 via Ollama, FAISS Vector DB, and sentence-transformers.
PhantomText: Hidden Text Steganography
Overview: PhantomText is a fully client-side web application that combines cryptography and steganography to hide encrypted messages inside ordinary-looking text using zero-width Unicode characters.
The Build: The application encrypts secret messages with AES-256, converts them into invisible zero-width characters, and embeds them inside normal text directly in the browser without sending data to any server.
Tech Stack: HTML5, CSS3, JavaScript, Web Crypto API, AES-GCM, PBKDF2, SHA-256, CompressionStream API
Handwritten Digit Recognition
Overview: A web application that allows users to upload an image of a handwritten digit (0–9) and predicts the number with high accuracy.
The Build: Developed and trained a Convolutional Neural Network (CNN) using the MNIST dataset.
Tech Stack: Python, Flask, OpenCV, and Scikit-Learn.
ICU Readmission Risk Predictor
Overview: Machine learning pipeline for predicting ICU readmission risk using patient data.
The Build: The application is built with Python, Scikit‑Learn, and Streamlit and deployed online for interactive predictions.
Tech Stack: Python, Pandas, NumPy, Scikit-Learn, Streamlit, Joblib, Google Drive (for model hosting)
Google Sites
Report abuse