DataEngBytes 2023 - BRI-T1-07 - Joe Reis
summary
Argues that data modeling is both abandoned and indispensable for AI, citing a real-world symptom of 39,000 DBT models in one company and widespread ad hoc, query-driven schemas. Points to missing conceptual and logical models, overreliance on physical implementations, and tool-heavy but practice-poor teams. Recommends modeling across the data life cycle, using reusable pattern libraries, data contracts and streaming feedback loops so models remain coherent, believable and usable for downstream ML and LLM deployments.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:11Data modeling matters now
- 2:59AI’s rise and data chaos
- 4:02Defining data modeling
- 7:14Data modeling feels dead
- 11:25Ignoring modeling consequences
- 13:55Hype, tools, and dev-data divide
- 16:00Bringing modeling back
- 17:45Modeling across lifecycle
- 19:30Ending modeling infighting
- 20:56Conclusion and future
- 21:37Q&A
- 22:39Q&A: Vendor myths
- 24:45Q&A: Tools vs modeling
related talks
26mDataEngBytes 2023 - MEL-T1-01 - Joe Reis
30mDATA MODELLING IS DEAD! LONG LIVE DATA MODELLING!
33mDataEngBytes 2023 - MEL-T1-02 - Kendra Vant
36mDataEngBytes 2023 - PER-T1-02 - Josh Armitage
45mMixed Model Arts
39mPanel Discussion
21mDataEngBytes 2023 - SYD-T2-04 - Suneeta Mall
01h 01mYou don't need AI; You need to know what your data is.
