BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20270314T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20261101T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260927T162346Z UID:E87E6201-4DBE-4F2A-924A-E9B3E104FEB9 DTSTART;TZID=America/New_York:20261109T180000 DTEND;TZID=America/New_York:20261109T193000 DESCRIPTION:[][Nutan_140x140]Nutan Sahoo\n\n[Advitya_140x140]\nAdvitya Gema wat\nDescription: Most standard RAG tutorials rely on a naive pattern: spl itting text\, embedding chunks\, and injecting the top matches directly in to an LLM prompt. While simple to implement\, this approach frequently fai ls in real-world scenarios\, retrieving semantically related but factually irrelevant context\, generating unverified assertions\, etc. In this hand s-on workshop\, we move beyond naive RAG to build a robust\, production-re ady RAG pipeline. Participants will learn how to design resilient architec tures that handle complex\, domain-specific document.\n\nKey Topics & Impl ementations Covered:\n\n- Structural Document Preprocessing & Chunking: Pa rsing hierarchical documents into semantically coherent units rather than arbitrary fixed-character splits.\n- Hybrid Search & Retrieval: Combining sparse keyword matching (BM25) with dense semantic embeddings to capture b oth exact terminology and conceptual intent.\n- Adaptive Feedback Learning : Integrating dynamic feedback for improving retrieval.\n- Sentence-Level In-Line Citations: Adding verifiable citation tags\, so every fact links b ack to its source document.\n- Cross-Session Persistence: Extracting and s toring high-signal user profile facts to maintain continuity across distin ct user sessions.\n\nCo-sponsored by: IEEE Boston AI Group\n\nSpeaker(s): Nutan Sahoo\, Advitya Gemawat\n\nRoom: NERD/1250\, Abigail Adams\, Microso ft NERD Center\, Kendall Square\, 1 Memorial Drive\, Cambridge\, Massachus etts\, United States\, 02142 LOCATION:Room: NERD/1250\, Abigail Adams\, Microsoft NERD Center\, Kendall Square\, 1 Memorial Drive\, Cambridge\, Massachusetts\, United States\, 02 142 ORGANIZER:info@voptronix.com SEQUENCE:87 SUMMARY:Building a Production-Grade Retrieval Augmented Generation (RAG) Pi peline URL;VALUE=URI:https://events.vtools.ieee.org/m/575323 X-ALT-DESC:Description: <br /><table style="border-collapse: collapse\; wid th: 93.5867%\; height: 519px\; border-width: 0px\;" border="1"><colgroup>< col style="width: 20.2037%\;"><col style="width: 79.7984%\;"></colgroup>\n <tbody>\n<tr style="height: 152.8px\;">\n<td style="text-align: center\; h eight: 152.8px\; border-width: 0px\;">\n<p><img src="https://events.vtools .ieee.org/vtools.ui/media/display/c29a0ee8-765e-42b9-99ec-75bbf1cff03c" al t=""><img style="display: block\; margin-left: auto\; margin-right: auto\; " src="https://events.vtools.ieee.org/vtools_ui/media/display/f37ceeb9-d1f 3-4349-ba56-90a3ee88b22c" alt="Nutan_140x140" width="140" height="140"><sp an style="font-size: 10pt\;"><strong>Nutan Sahoo</strong></span></p>\n<p>< span style="font-size: 10pt\;"><strong><img src="https://events.vtools.iee e.org/vtools_ui/media/display/c0f0e4ef-8d7f-46a4-8972-5d1c03985c9c" alt="A dvitya_140x140" width="140" height="140"><br>Advitya Gemawat</strong></spa n></p>\n</td>\n<td style="height: 152.8px\; border-width: 0px\;"><strong>D escription:&nbsp\;</strong>Most standard RAG tutorials rely on a naive pat tern: splitting text\, embedding chunks\, and injecting the top matches di rectly into an LLM prompt. While simple to implement\, this approach frequ ently fails in real-world scenarios\, retrieving semantically related but factually irrelevant context\, generating unverified assertions\, etc. In this hands-on workshop\, we move beyond naive RAG to build a robust\, prod uction-ready RAG pipeline. Participants will learn how to design resilient architectures that handle complex\, domain-specific document.<br>\n<p cla ss="MsoNormal" style="line-height: 2\;"><strong>Key Topics &amp\; Implemen tations Covered:</strong></p>\n<ul style="margin-top: 0in\;" type="disc">\ n<li class="MsoNormal" style="mso-list: l0 level1 lfo1\; tab-stops: list . 5in\;"><strong>Structural Document Preprocessing &amp\; Chunking:</strong> &nbsp\;Parsing hierarchical documents into semantically coherent units rat her than arbitrary fixed-character splits.</li>\n<li class="MsoNormal" sty le="mso-list: l0 level1 lfo1\; tab-stops: list .5in\;"><strong>Hybrid Sear ch &amp\; Retrieval:</strong>&nbsp\;Combining sparse keyword matching (BM2 5) with dense semantic embeddings to capture both exact terminology and co nceptual intent.</li>\n<li class="MsoNormal" style="mso-list: l0 level1 lf o1\; tab-stops: list .5in\;"><strong>Adaptive Feedback Learning:</strong>& nbsp\;Integrating dynamic feedback for improving retrieval.</li>\n<li clas s="MsoNormal" style="mso-list: l0 level1 lfo1\; tab-stops: list .5in\;"><s trong>Sentence-Level In-Line Citations:</strong>&nbsp\;Adding verifiable c itation tags\, so every fact links back to its source document.</li>\n<li class="MsoNormal" style="mso-list: l0 level1 lfo1\; tab-stops: list .5in\; "><strong>Cross-Session Persistence:</strong>&nbsp\;Extracting and storing high-signal user profile facts to maintain continuity across distinct use r sessions.</li>\n</ul>\n</td>\n</tr>\n</tbody>\n</table>\n<p class="MsoNo rmal">&nbsp\;</p> END:VEVENT END:VCALENDAR