The Analytics Engineering Podcast Inside Snowflake's AI roadmap (w/ Chris Child)
103 snips
Dec 14, 2025 Chris Child, VP of Product Management at Snowflake, dives into the future of data engineering and AI. He highlights the evolution from Snowpark to Cortex and the importance of row-column governance for AI agents. Chris discusses Snowflake's commitment to Apache Iceberg for better interoperability and how artificial intelligence is reshaping data access and processing. He predicts a shift towards standardized data products and emphasizes embedding semantic context in data workflows for more intelligent decision-making.
AI Snips
Chapters
Transcript
Episode notes
Run Code Inside The Data Boundary
- Snowflake built Snowpark to run non‑SQL workloads inside the platform to avoid customers pulling data out for ML and other tasks.
- Early adopters used it for some workloads but training most ML models still happens outside Snowflake.
Acquisitions Brought LLM Features In‑House
- Snowflake acquired Applica for document extraction and Neva for LLM-based web search to bring unstructured capabilities in-house.
- These acquisitions helped Snowflake offer extraction from PDFs and better enterprise information retrieval.
LLMs Are Strong At Text And Code, Weak At SQL
- LLMs excel at unstructured text but struggle with structured business queries and deterministic SQL generation.
- They are already helpful at generating code and dbt artifacts to accelerate engineering work.