BEGIN:VCALENDAR VERSION:2.0 PRODID:-//CERN//INDICO//EN BEGIN:VEVENT SUMMARY:Deployment considerations and wrap-up DTSTART;VALUE=DATE-TIME:20260707T120000Z DTEND;VALUE=DATE-TIME:20260707T121500Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1157@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\, Sotiris Gyftopoulos (ATHENA RC)\nThis wrap-up session discusses how a Colab-based RAG prototyp e can be prepared for real-world deployment. Participants will review the main components of a deployed RAG application\, including user interface\, API or orchestration chain\, retriever service\, vector or hybrid store\, LLM service\, monitoring and logging. The session highlights production c oncerns such as latency\, cost\, data privacy\, access control\, source up dates\, model versioning\, failure logging and regression testing. It also revisits the specific challenges of Greek-language public-service RAG\, i ncluding morphology\, administrative terminology\, tables\, articles\, cod es\, appendices and vague user questions. Participants will receive an end -to-end design checklist covering information need\, risk level\, document preparation\, chunking\, metadata\, retrieval strategy\, grounding\, cita tions\, evaluation and deployment safeguards. The session closes the webin ar by connecting the theoretical concepts and hands-on notebook into a coh erent pathway for operational RAG applications. It also identifies next st eps for adapting the tutorial to participants' own organisational domains and use cases.\n\nhttps://events.grnet.gr/event/213/contributions/1157/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1157/ END:VEVENT BEGIN:VEVENT SUMMARY:RAG evaluation: retrieval quality\, faithfulness\, groundedness an d RAGAS/DeepEval concepts DTSTART;VALUE=DATE-TIME:20260707T114000Z DTEND;VALUE=DATE-TIME:20260707T120000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1156@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\, Sotiris Gyftopoulos (ATHENA RC)\nThis evaluation session explains how to assess RAG systems be yond a single final answer score. Participants will learn why evaluation m ust decompose the pipeline into retrieval\, generation\, citation quality and end-to-end behaviour. The session introduces retrieval metrics such as Precision@k\, Recall@k\, Mean Reciprocal Rank and NDCG\, showing how they describe the evidence made available to the model. It then discusses answ er-level criteria\, including correctness\, faithfulness\, groundedness\, citation quality\, completeness and relevance. Participants will also be i ntroduced to RAGAS and DeepEval concepts for automated RAG evaluation\, re gression testing and structured comparison of system variants. The emphasi s is diagnostic: metrics should identify failure modes and guide concrete improvements\, such as better chunking\, query handling\, filtering\, rera nking or prompt constraints. By the end\, participants will understand how to evaluate and iterate RAG systems systematically. This prepares them to maintain RAG quality as data\, prompts and models evolve in operational s ettings\, after deployment too.\n\nhttps://events.grnet.gr/event/213/contr ibutions/1156/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1156/ END:VEVENT BEGIN:VEVENT SUMMARY:Hands-on 4: end-to-end RAG answer generation with source attributi on DTSTART;VALUE=DATE-TIME:20260707T112000Z DTEND;VALUE=DATE-TIME:20260707T114000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1155@events.grnet.gr DESCRIPTION:Speakers: Sotiris Gyftopoulos (ATHENA RC)\nThis fourth hands- on tutorial connects retrieval and generation into an end-to-end RAG chain . Participants will start from a user question\, retrieve relevant chunks\ , build a grounded prompt\, call the selected language model and return an answer with source attribution. The notebook will expose each intermediat e step so participants can see exactly what the model receives before gene rating a response. They will compare answers produced with and without ret rieved context\, highlighting the role of evidence in reducing hallucinati ons and improving factuality. The session also demonstrates how to handle insufficient evidence by producing cautious responses rather than unsuppor ted claims. Participants will inspect whether cited sources actually suppo rt the generated answer and prepare the chain for a simple user-facing int erface. By the end\, participants will have a runnable RAG workflow that l inks retrieval\, prompt construction\, generation and transparent citation s. This workflow provides practical foundations for later evaluation and d eployment discussions in the webinar.\n\nhttps://events.grnet.gr/event/213 /contributions/1155/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1155/ END:VEVENT BEGIN:VEVENT SUMMARY:Hands-on 3: semantic search\, BM25/hybrid retrieval and reranking DTSTART;VALUE=DATE-TIME:20260707T103000Z DTEND;VALUE=DATE-TIME:20260707T105500Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1153@events.grnet.gr DESCRIPTION:Speakers: Sotiris Gyftopoulos (ATHENA RC)\nThis third hands-o n tutorial lets participants compare retrieval strategies directly in the Colab notebook. Using representative Greek-language public-service questio ns\, participants will run semantic search\, keyword-based search and hybr id retrieval over the same indexed dataset. They will inspect returned chu nks\, compare ranking behaviour and identify cases where one retrieval str ategy succeeds while another misses important evidence. The session also i ntroduces practical analysis of false positives\, missing evidence and noi sy context. Participants will apply reranking to candidate chunks and comp are the Top-K context before and after reranking\, observing how ordering affects the evidence available to the generator. The tutorial encourages p articipants not to trust retrieval scores blindly\, but to read and evalua te the retrieved passages. By the end\, participants will understand how t o choose retrieval strategies for different query types and how to diagnos e retrieval weaknesses systematically. These observations support later gr ounding\, citation checking and end-to-end answer evaluation activities in the notebook.\n\nhttps://events.grnet.gr/event/213/contributions/1153/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1153/ END:VEVENT BEGIN:VEVENT SUMMARY:Generation in RAG: context construction\, grounding\, citations an d answer constraints DTSTART;VALUE=DATE-TIME:20260707T105500Z DTEND;VALUE=DATE-TIME:20260707T112000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1154@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\nThis methodology sessi on explains how retrieved evidence is transformed into a grounded answer t hrough prompt construction and generation. Participants will learn the mai n components of a grounded RAG prompt: the user query\, retrieved evidence \, system instructions and required answer format. The session discusses h ow prompt instructions can constrain the model to answer only from availab le sources\, cite supporting evidence\, avoid unsupported claims and state when the retrieved context is insufficient. It also covers context constr uction decisions\, including how much evidence to include\, how to order p assages and how to separate facts from caveats. Special attention is given to source attribution and citation granularity\, particularly in public-s ervice or administrative domains where users need traceability. By the end \, participants will understand that generation quality depends not only o n the LLM\, but also on retrieval quality\, prompt design and evidence org anisation. These principles are then implemented in the following hands-on end-to-end RAG tutorial.\n\nhttps://events.grnet.gr/event/213/contributio ns/1154/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1154/ END:VEVENT BEGIN:VEVENT SUMMARY:Retrieval strategies: semantic\, lexical\, hybrid retrieval\, HyDE \, RRF and reranking DTSTART;VALUE=DATE-TIME:20260707T100500Z DTEND;VALUE=DATE-TIME:20260707T103000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1152@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\nThis methodology sessi on examines the retrieval stage\, which determines what evidence the langu age model can use when generating an answer. It compares dense semantic re trieval\, sparse keyword retrieval\, structured retrieval and hybrid retri eval\, explaining when each modality is useful. Participants will see why semantic search is valuable for paraphrases and user-friendly language\, w hile keyword or BM25-style search is important for exact administrative te rms\, identifiers\, article numbers\, dates and rare phrases. The session also introduces query enhancement techniques such as rewriting\, expansion and HyDE\, which can bridge the gap between conversational user queries a nd formal source language. It then explains fusion\, Reciprocal Rank Fusio n and reranking as ways to combine complementary retrieval signals and imp rove the final Top-K evidence. Participants will understand how retrieval choices affect recall\, precision\, context noise\, latency and final answ er reliability. The session prepares participants to experiment with these choices in tutorial exercises using realistic Greek queries.\n\nhttps://e vents.grnet.gr/event/213/contributions/1152/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1152/ END:VEVENT BEGIN:VEVENT SUMMARY:Hands-on 2: chunking\, metadata and vector indexing DTSTART;VALUE=DATE-TIME:20260707T093500Z DTEND;VALUE=DATE-TIME:20260707T095000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1151@events.grnet.gr DESCRIPTION:Speakers: Sotiris Gyftopoulos (ATHENA RC)\nThis second hands- on tutorial applies the knowledge preparation concepts introduced in the m ethodology session. Participants will create document chunks from the load ed public-service dataset\, attach useful metadata and generate embeddings for each chunk. The session demonstrates how metadata such as document ti tle\, section\, source identifier and chunk position helps preserve tracea bility and makes later retrieval more controllable. Participants will then build a local vector index in the Colab environment and run an initial si milarity search to inspect the retrieved evidence. The emphasis is on visi bility and debugging: participants will read example chunks\, check whethe r they are coherent\, verify that metadata is correct and examine whether the first retrieval results are meaningful. By the end\, participants will have a working indexed knowledge base and will understand how data prepar ation decisions shape everything that follows in the RAG pipeline. This re sult becomes the basis for hybrid retrieval\, reranking and answer generat ion exercises.\n\nhttps://events.grnet.gr/event/213/contributions/1151/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1151/ END:VEVENT BEGIN:VEVENT SUMMARY:Ingestion\, chunking\, metadata and embeddings for RAG DTSTART;VALUE=DATE-TIME:20260707T091000Z DTEND;VALUE=DATE-TIME:20260707T093500Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1150@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\nThis methodology sessi on focuses on preparing knowledge so that it can be retrieved accurately b y a RAG system. It explains why document preparation is not a minor prepro cessing task but a core design decision that affects downstream retrieval and answer quality. Participants will learn how text can be extracted from PDFs\, Markdown\, HTML or structured files while preserving headings\, se ctions\, tables\, article numbers\, source identifiers and other traceable information. The session then introduces chunking strategies\, including fixed-size\, overlapping\, semantic and structure-aware chunking\, and dis cusses how chunk size influences retrieval precision and context completen ess. It also covers metadata fields that support filtering\, citation and source attribution. Finally\, the session explains embeddings and vector i ndexing\, with emphasis on multilingual or Greek-aware representations\, i ndex choice\, latency\, scaling\, filtering and maintainability in prototy pe and production RAG settings. These foundations prepare participants for the following indexing and retrieval implementation exercises in Colab\, including practical debugging.\n\nhttps://events.grnet.gr/event/213/contri butions/1150/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1150/ END:VEVENT BEGIN:VEVENT SUMMARY:Hands-on 1: dataset loading\, document preparation and first RAG p ipeline overview DTSTART;VALUE=DATE-TIME:20260707T084000Z DTEND;VALUE=DATE-TIME:20260707T091000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1149@events.grnet.gr DESCRIPTION:Speakers: Sotiris Gyftopoulos (ATHENA RC)\nThis first hands-o n tutorial introduces the Google Colab notebook and the running example us ed throughout the practical part of the webinar. Participants will load a small Greek-language public-service dataset and inspect how documents are represented before indexing. The session focuses on orientation rather tha n optimisation: understanding the structure of the corpus\, identifying th e available fields\, examining source information and seeing how raw docum ents become inputs to a RAG pipeline. Participants will review the full no tebook flow\, from setup and data loading to chunking\, embeddings\, retri eval\, answer generation and evaluation. They will also inspect intermedia te outputs so that the pipeline remains transparent and debuggable. By the end of the session\, participants will understand the dataset\, the role of metadata\, the overall sequence of implementation steps and how the not ebook will be used to build a practical RAG system progressively. This fou ndation helps them follow later implementation choices with confidence dur ing exercises.\n\nhttps://events.grnet.gr/event/213/contributions/1149/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1149/ END:VEVENT BEGIN:VEVENT SUMMARY:RAG architectures: Naïve\, Advanced and Modular DTSTART;VALUE=DATE-TIME:20260707T081500Z DTEND;VALUE=DATE-TIME:20260707T084000Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1140@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\nThis theory session pr esents the main architectural paradigms used in Retrieval-Augmented Genera tion systems. It starts with Naïve RAG as the baseline pipeline: document ingestion\, chunking\, embedding generation\, vector indexing\, retrieval and grounded answer generation. It then explains how Advanced RAG improve s this baseline through better preprocessing\, metadata enrichment\, query rewriting\, hybrid retrieval\, reranking\, context compression and eviden ce filtering. The session also introduces Modular RAG\, where routing\, re writing\, retrieval\, fusion\, generation and verification can be organise d as configurable components rather than a fixed linear chain. Participant s will learn how each architecture changes the quality\, reliability\, lat ency and maintainability of a RAG application. The session prepares partic ipants to recognise which design pattern is appropriate for a simple educa tional prototype\, a more robust domain assistant or a production system t hat must adapt to different question types and knowledge sources. It also clarifies how the same concepts appear later in the hands-on notebook exer cises.\n\nhttps://events.grnet.gr/event/213/contributions/1140/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1140/ END:VEVENT BEGIN:VEVENT SUMMARY:Why RAG? Hallucinations\, knowledge cut-offs and grounded generati on DTSTART;VALUE=DATE-TIME:20260707T080000Z DTEND;VALUE=DATE-TIME:20260707T081500Z DTSTAMP;VALUE=DATE-TIME:20260702T234219Z UID:indico-contribution-213-1139@events.grnet.gr DESCRIPTION:Speakers: George Drosatos (ATHENA RC)\nThis opening session e xplains why Retrieval-Augmented Generation is needed when Large Language M odels must answer with factual\, current and verifiable information. It in troduces the main limitations of standalone LLMs\, including static traini ng knowledge\, knowledge cut-offs\, hallucinations and weak handling of sp ecialised domain terminology. Using Greek public-service information as a motivating example\, the session shows how RAG changes the task from gener ating from model memory to answering from retrieved evidence. Participants will see how an external knowledge source can support more transparent\, context-aware responses\, especially in domains where users need reliable guidance\, citations and clear caveats. The session establishes the centra l concepts used throughout the webinar: user query\, retrieval\, evidence\ , grounded answer and source attribution. By the end\, participants will u nderstand the practical motivation for RAG and why retrieval quality matte rs before any answer is generated by the model. It clearly frames RAG as a reliability pattern for specialised knowledge-intensive applications.\n\n https://events.grnet.gr/event/213/contributions/1139/ LOCATION: URL:https://events.grnet.gr/event/213/contributions/1139/ END:VEVENT END:VCALENDAR