Adrian Brudaru | Data Track | Compass AI & Data Summit 2025
summary
Argues that an LLM-native workflow can automate ELT by having models read docs, fill configuration templates, and generate connector scaffolds that call REST APIs as Python dicts; uses DT/DT hub, cursor, DLT, ibis, Arrow and Marimo notebooks plus a pipeline dashboard for validation. Reports: 5,000 production users, a public backlog near 6,000 connectors, target of 10k LLM contexts, agency users saying ~50% dev time saved and pipelines running in under an hour, claiming extract-and-load is largely solved.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:09LLM data engineer rise
- 3:57Job evolution pace
- 7:29LLM-native workflow demo
- 12:13Automating ELT at scale
- 15:31Data engineer’s future role
- 19:05Iterating LLM-native workflows
- 28:55Q&A on runtimes
- 36:05Closing remarks
related talks
10mSteering Semantic Data Processing With DocWrangler
22mCase Study in Bridging Production Software and Data Practices for LLM Model Training
33mDesigning LLM Applications for Massive Datasets
32mDocuments Meet LLMs: Tales from the Trenches
24mDemocratizing data access through LLM
30mWhat Can LLMs Do with Messy Data?
16mDemocratizing Data Pipelines: Introducing the AI-Powered ETL Assistant
49mNext-Gen Documentation with AI