RAG-Fusion: multi-query generation + Reciprocal Rank Fusion for better retrieval-augmented generation. Includes evaluation harness with NFCorpus/BEIR.

Two-tier hybrid search for Rust: sub-millisecond initial results via potion-128M, quality-refined rankings in 150ms via MiniLM-L6-v2. Combines lexical (Tantivy BM25) and semantic (vector cosine) search with Reciprocal Rank Fusion. Progressive iterator API, f16 SIMD vector index, feature-gated compilation.

Performance Evaluation of Rankers and RRF Techniques for Retrieval Pipelines: Employs Diversity, Lost-in-the-Middle, and Similarity rankers to reorder documents and maximize LLM context window performance. Implements Hybrid Retrieval with Reciprocal Rank Fusion (RRF) and rigorous BEIR evaluation (NDCG, MAP, Recall, Precision).

Allows you to merge search results from multiple search engines using the reciprocal rank fusion algorithm.

A state-of-the-art Information Retrieval system for TREC ROBUST04 achieving MAP 0.3309. Features a Novel 4-Way RRF Fusion architecture combining BM25, Neural Reranking (Cross-Encoders), and LLM-Augmented Query Expansion (Query2Doc) via LiteLLM.

Analysis and implementational details of rank aggregation methods for combing results of multiple engines to achieve best mAP and P@5,10 values

This demo showcases a store associate application built on MongoDB Atlas, created to streamline product discovery and inventory visibility as part of a unified commerce strategy.

Production RAG system for international student tax advisory. Hybrid Elasticsearch+BM25 retrieval, LangChain LCEL, LLaMA/Gemini routing, RAGAS evaluation (GPT-4 judge), Prometheus/Grafana observability. Deployed on Render with Supabase logging.

Hybrid document fusion multi-step retrieval for multi-modal RAG

Hybrid RAG with three retrievers—Lexical (BM25), Semantic (embeddings), and Hybrid (Reciprocal Rank Fusion) — parallel retrieval, Llama 3 generation, and side-by-side evaluation with reproducible notebooks.

A local-first RAG demo that emphasizes production-grade patterns: deterministic SHA256 IDs for idempotent ingestion, hybrid retrieval (Chroma + BM25) fused with RRF, HMAC-signed BM25 indices, input validation + resource limits, structured tracing, and practical runtime optimizations (RAGService reuse, embedding cache, context budgets).

A high-performance Retrieval-Augmented Generation pipeline for technical Q&A workloads. Combines hybrid retrieval (dense + BM25), query expansion, Reciprocal Rank Fusion (RRF), and cross-encoder re-ranking to improve retrieval precision and answer grounding. Evaluated with Ragas, showing measurable gains in context recall and faithfulness.

Repository for the CENG543 Course Information Retrieval Term Project

Hybrid retrieval engine: PGVector dense + BM25 sparse search with Reciprocal Rank Fusion and dynamic query routing

Hybrid Retrieval-Augmented Generation (RAG) system for SME customer support, combining BM25 sparse retrieval and Pinecone dense retrieval via Reciprocal Rank Fusion to answer technical queries grounded in company documentation. Built for Benamdaj Dredging Solutions Ltd. and evaluated with RAGAS across four ablation configurations.

LangChain vector store that combines BigQuery VECTOR_SEARCH and SEARCH for hybrid semantic + keyword retrieval

Production-grade hybrid RAG: BM25 + dense retrieval + cross-encoder reranking + LLM-as-Judge

An AI-powered public grievance and complaint management platform built on the MERN stack. Features automated LLM-based priority routing, semantic search, and real-time dashboards for citizens and administrators.

RAG pipeline over real Amazon Q&A data - hybrid retrieval, LLM evaluation, streaming UI, Docker and CI.

Hybrid RAG AI chat app for querying documents using vector search, keyword search, and LLMs.