Marrying Structured + Unstructured Data
summary
Claims a majority of enterprise data is untouched and shows concrete wins: extracting transcripts into a lakehouse reduces manual review, and vision models find manufacturing defects that ERP tables miss. Implementation details include prototyping on AWS Glue + Lambda + Textract and converting images to CSV/JSON for ERP ingestion. Vision stack uses YOLO and SAM with multimodal LLMs for visual question answering. The talk reports a striking 80% unstructured data figure, live prototypes, and plans for closed‑loop defect prevention.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:48Structured vs unstructured data
- 7:25Bridging data in manufacturing
- 11:46Insights from shop floor data
- 16:24Vision AI evolution
- 18:11AI in higher education admissions
- 24:50Key considerations
- 25:32Future of computer vision
related talks
30mMastering Unstructured Data with AI and ML
30mExtracting insights from unstructured data
46mBusiness: AI as an enabler for unlocking Industry Data: Monetizing Enterprise Value
08mLeveraging data insights to increase efficiency and innovation
01h 00mOpening Keynote
29mData management for enterprise-grade foundation models.
31mCombining NLP with Structured Data
16mBeyond ETL: Revolutionizing Data Engineering Workflows