BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:SmartAttica Τraining modules for SMEs - Module 13 "Retrieval-Augm
 ented Generation (RAG) with Docling Document Parsing"
DTSTART;VALUE=DATE-TIME:20260624T080000Z
DTEND;VALUE=DATE-TIME:20260624T110000Z
DTSTAMP;VALUE=DATE-TIME:20260612T223823Z
UID:indico-event-210@events.grnet.gr
DESCRIPTION:\n\n\n\nGRNET announces\, in the context of SmartAttica EDIH (
 European Digital Innovation Hub)\, the 13th Module of Τraining modules fo
 r SMEs  with the subject "Retrieval-Augmented Generation (RAG) with Doclin
 g Document Parsing". \n\nDate: June 24th\, 2026\n\nLocation: Online via 
 Zoom \n\nPresentation Languages: Greek\, English\n\nInstructors: Nikos Ba
 kas (GRNET)\, Roman Dolgopolyi (GRNET)\n\nDuration: 3 hours\n\nDescription
 : This is a hands-on introduction to turning real-world documents into kno
 wledge a language model can use. Real documents are not plain text - they 
 contain headings\, tables\, figures\, captions\, and reading order. Partic
 ipants learn how Docling converts a PDF into structured content using layo
 ut-aware vision models\, how to inspect and chunk that content intelligent
 ly\, and how to assemble a complete Retrieval-Augmented Generation (RAG) p
 ipeline: chunk\, embed\, retrieve\, and generate a grounded\, citable answ
 er.\n\nTarget Audience: This module is designed for SME developers\, techn
 ical leads\, and data scientists who want to build question-answering syst
 ems over their own documents. It is ideal for those looking to ground lang
 uage models in trusted internal knowledge sources.\n\nLearning Objectives:
 \n\nBy the end of this module\, participants will be able to:\n\n\n	Conver
 t PDFs into structured documents with Docling and inspect the resulting do
 cument tree.\n	Extract and work with text blocks\, tables\, figures\, and 
 captions.\n	Split documents into retrieval-ready chunks while preserving s
 tructure and context.\n	Embed chunks and retrieve the most relevant ones u
 sing semantic similarity.\n	Assemble a RAG prompt and generate an answer t
 hat cites its sources.\n\n\nPrerequisites:\n\nParticipants should have:\n\
 n\n	Basic understanding of Python programming.\n	Familiarity with embeddin
 gs and semantic similarity.\n	Interest in NLP and document-processing appl
 ications.\n	Some experience with machine learning will be helpful.\n\n\nIn
 dicative Content:\n\n\n	Why Document Structure Matters. The gap between ra
 w PDF text and structured\, machine-usable content.\n	Converting PDFs with
  Docling. Layout detection\, reading order\, and the structured document t
 ree.\n	Inspecting the Document. Exploring texts\, groups\, figures\, and c
 aptions\; previewing as Markdown.\n	Working with Tables. Recovering rows\,
  columns\, and headers and exporting tables to Markdown.\n	Chunking for RA
 G. Structure-aware splitting that keeps headings and context attached to e
 ach chunk.\n	Embedding the Chunks. Turning chunks into vectors with Senten
 ce Transformers.\n	Retrieval. Embedding the query and finding the most rel
 evant chunks via cosine similarity.\n	Building the Prompt. Assembling labe
 lled context and instructing the model to cite section and page.\n	Generat
 ing the Answer. Passing the retrieved context to an LLM for a grounded res
 ponse.\n	Summary and Q&A. Key takeaways and open discussion.\n\n\n \n\nTh
 e project is co-funded by the European Union. Views and opinions expressed
  are however those of the author(s) only and do not necessarily reflect th
 ose of the European Union or the European Commission. Neither the European
  Union nor the granting authority can be held responsible for them. \n\nht
 tps://events.grnet.gr/event/210/
LOCATION:
URL:https://events.grnet.gr/event/210/
END:VEVENT
END:VCALENDAR