multilingual-e5

Fully offline RAG for Android: PDF ingestion, on-device embeddings (ONNX multilingual-e5-small) and grounded chat with LFM2-1.2B-RAG via the LEAP SDK. Kotlin, Jetpack Compose, Hilt, Room, Navigation 3.

A reproducible benchmark for incomplete-cue personal mobility recall, comparing BM25, multilingual E5 dense retrieval, and dense-seeded graph retrieval with RRF/PPR diagnostics.

Semantic retrieval pipeline for the Eastern Slavic protest-video corpus (v2)

2026 Natural Language Processing Course at FRI - Development of RAG for FRI students

This project extends the v2-0 Hybrid Model (statistical features + TTM embeddings) by adding a third modality: text embeddings derived from equipment master data. The three feature vectors — statistical x ∈ ℝ²⁸, TTM embedding y ∈ ℝ⁶⁴, and text embedding z ∈ ℝ¹⁰²⁴ — are concatenated into a 1,116-dimensional triplet feature h and fed into a LightGBM.

VoxRAG is a voice- and text-enabled Retrieval-Augmented Generation system built on MSMARCO-XI.

Infer a GitHub user's country from their public profile, offline in Node or the browser. Fine-tuned multilingual-e5-small (ONNX int8, 45 MB, included): 98.8% accuracy on users with a location. Live demo maps the authors of GitHub's latest events.

An experimental project designed for embedding-based text similarity search in web pages.

A production-grade multilingual RAG system for Moroccan legal question answering, grounded in official Arabic legal texts with traceable citations and a systematic evaluation framework.