Built at Eurac Research · 2026
Pipeline for converting multilingual books (PDF/EPUB) to structured Markdown via VLM inference with Qwen3-VL. Includes metadata extraction, quality evaluation (NED, BLEU, MarkdownStructureF1, BERTScore) and dependency parsing with Stanza.
Personal project
AI agent that searches academic job boards (Euraxess, MLScientist, jobs.ac.uk), scores positions against a CV, and generates tailored cover letters. 100% free, runs on Groq or Ollama.
For experts
Built at Compet-e Srl · 2025
Evaluation framework for machine translation quality on multilingual GDPR legal documents. Benchmarks local LLMs via Ollama with COMET, punctuation consistency, and length ratio metrics.
For experts
Personal project
Multilingual RAG chatbot (IT/EN/ES/FR/DE) for ELAN annotation software. Uses Qdrant vector search, Sentence Transformers, and Llama 3.3 70B via HuggingFace Inference API.
For experts
Built at Università di Napoli L'Orientale · 2024–2025
Context-specific chatbot that guides visitors through the history and culture of Naples. Built with BerkeleyDB, Sentence Transformers, and Llama 3.2 via llama.cpp for local inference.
For experts