{ "$schema": "/api/schemas/documents.schema.json", "apiVersion": "1.0.0", "generatedAt": "2026-07-14T19:46:21.019Z", "sourcePath": "src/main/data/docs", "docId": "978-1-61482-966-9-28", "document": { "abstract": "Since 2015, Trinity College Dublin has delivered a Master's-level “Motion Picture Engineering” (MPE) course uniquely combining core principles of video compositing and streaming with cutting-edge developments in the field. In 2025, to address diverse student backgrounds - from deep learning practitioners to image processing newcomers - we developed a domain-specific AI Teaching Assistant (AI-TA) using Retrieval Augmented Generation (RAG). While RAG is a well-established paradigm, this paper addresses practical implementation challenges and classroom integration constraints in real-world educational deployment. Our technical overview includes details about architecture, security, compliance, document embedding, vector store tuning, prompt engineering and UI integration. The AI-TA was deployed to 43 MPE students over 7 weeks during Spring 2025. Anonymous telemetry provided insight into students' engagement patterns. We further leveraged AI for post-deployment exploration of student queries with the goal of identifying commonly misunderstood topics and knowledge gaps, enabling data-informed curriculum refinement. Assessment outcomes showed no significant exam performance differences across varying AI-TA usage levels, indicating that with thoughtfully designed open-book assessments, AI tools may enhance learning without compromising academic integrity. Our case study, however, reveals “muddy waters” in practical AI-TA implementation, from questioning the benefits of student anonymity to management of an operationally intensive lecturer's workflow. Student feedback revealed that they found the AI-TA useful for clarifying and exploring MPE topics but remained cautious about its use as a substitute for human tutoring and skeptical of its suitability in open-book exams. This work provides practical guidance for AI-TA deployment, addressing real-world implementation challenges applicable in MPE and broader technical education contexts. Our code is available on GitHub 1 .", "abstract$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "authors": [ { "name": "Deirdre O'Regan" }, { "name": "Anil C. Kokaram" } ], "authors$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "contentType": "orig-research", "contentType$meta": { "confidence": "high", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "copyright": { "holder": "SMPTE", "holder$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "year": "2025", "year$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" } }, "copyright$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "docId": "978-1-61482-966-9-28", "docId$meta": { "confidence": "high", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "docLabel": "MTS 2025, Article 28 (pp. 1 to 23)", "docLabel$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "docTitle": "Implementing an AI Teaching Assistant for a Class in Motion Picture Engineering", "docTitle$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "docType": "Conference Paper", "docType$meta": { "confidence": "high", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "isbn": { "electronic": "978-1-61482-966-9", "electronic$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" } }, "isbn$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "journalAcronym": "MTS", "journalAcronym$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "journalTitle": "SMPTE 2025 Media Technology Summit", "journalTitle$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "keywords": [ "Motion Picture Engineering", "Education", "Artificial Intelligence", "Retrieval Augmented Generation", "RAG", "Production" ], "keywords$meta": { "confidence": "high", "note": "Long-tail keyword cleanup (keywordLongTailApply.js): prose/contentless terms dropped, acronym casing + typos fixed, duplicate variants folded; all surviving terms indexed in controlledKeywords.", "source": "resolved", "updated": "2026-07-14T16:29:38.680Z" }, "pages": "1–23", "pages$meta": { "confidence": "high", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "publicationDate": "2025-10-13", "publicationDate$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "publisher": "SMPTE", "publisher$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "publisherLocation": { "city": "Pasadena, CA", "city$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" } }, "publisherLocation$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" }, "status": { "active": true, "active$meta": { "confidence": "medium", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" } }, "volume": "00", "volume$meta": { "confidence": "high", "note": "Ingested from SMPTE canonical content_batch via ingestNlmCanonicalDocs.js", "source": "parsed", "updated": "2026-07-10T20:01:10.986Z" } } }