L

llm

Offline, CPU-only document to structured JSON extractor

MedRepo is an offline, CPU-first AI application designed to convert medical Complete Blood Count (CBC) reports from PDFs or images into clean, structured JSON. The project prioritizes patient privacy by processing all data locally without requiring an internet connection or cloud services. It combines OCR and lightweight local language models to accurately extract and organize medical information for further analysis or integration with healthcare applications. Built with a modular architecture, MedRepo is easy to set up, extend, and maintain. The project aims to make medical report digitization faster, more secure, and accessible on low-resource devices while supporting future enhancements and broader diagnostic report formats.

Local-first RAG platform with ChromaDB, Ollama, Groq BYOK, semantic retrieval, document intelligence, and business insights.

A Retrieval-Augmented Generation (RAG) chatbot built with Streamlit, LangChain, FAISS, and sentence-transformers. Upload PDFs and ask questions using Gemini or local Llama.cpp LLMs.

Fully offline document Q&A system using PyMuPDF4LLM, Ollama, and llama.cpp. Dual LLM backend with pre-commit quality gates.

Local-First Document Q&A with RAG using Ollama — Find-Retrieve-Answer pattern

A Streamlit-based research paper assistant that uses AI (RAG + PyMuPDF4LLM) for document summarization, Q&A, and multimedia analysis with local-first inference via Ollama.

A local-first Retrieval-Augmented Generation chatbot for PDF Q&A built with Streamlit, LangChain, ChromaDB, and llama.cpp

Mana Ruchulu is a Telugu-language recipe sharing and discovery platform built using Streamlit. It allows users to:

Create an account and log in securely

Upload recipes with text, images, videos, or audio instructions

View and interact with other users’ recipes

Participate in weekly cooking challenges

Post short-lived “stories” about their cooking experiences

Compete on a leaderboard based on points earned from contributions

The project supports both SQLite and Supabase backends, making it deployable on platforms like Hugging Face Spaces while being scalable for production use.

Its design emphasizes Telugu cultural heritage by supporting native script and traditional dish categorization.

A family recipe sharing app built with Python for managing and sharing recipes across family members.