hybrid-search-technique

A verifiable engineering knowledge engine that turns technical documents, CAD models, and equipment data into structured knowledge for humans and AI agents.

The purpose of this GitHub repository is to furnish the source code corresponding to the research paper authored by me. The repository is designed to facilitate the verification of results by users who wish to apply their own datasets.

AI-Spaces integrates React, Redux Toolkit, and FastAPI to build an advanced platform for interacting with AI models. Users can input questions or upload documents to receive precise, AI-driven summaries, creating a seamless and intuitive experience for modern web applications.

A notes app with hybrid recall — BM25 keyword search and semantic embeddings fused into one ranked result. FastAPI, Next.js, SQLite FTS5, Chroma.

API de RAG de alto rendimiento con FastAPI. Preguntas y respuestas sobre PDFs usando búsqueda híbrida (Qdrant + BM25), re-ranking y LLMs locales con Ollama.

Multilingual (Urdu/Roman Urdu/English) RAG system for Class 11 Islamiyat — hybrid retrieval (semantic + BM25), Cohere reranking, dual-source page citation, and hallucination-safe grounded answers from Textbook + Guide Book.

Structure-aware RAG pipeline for enterprise Confluence — hybrid retrieval, contextual chunking, cross-encoder reranking. Production system: Retrieval@5 lifted from 0.41→0.78, hallucination rate <1%, p95 latency <900ms.

Hybrid Search RAG Pipeline integrating BM25 and vector search techniques using LangChain

Training Data Generator for SPLADE Model Fine-tuning