This research identifies critical gaps in deploying LLMs for enterprise data engineering tasks, highlighting differences from academic benchmarks.

• Scale challenges: Enterprise tables often exceed academic dataset sizes by 10-100x
• Task complexity: Real-world data engineering requires more complex operations than standard benchmarks
• Internal knowledge: Enterprises need LLMs that integrate with proprietary domain knowledge
• Risk management: Production data engineering demands higher accuracy and reliability than current LLM capabilities

For engineering teams, this research provides a roadmap for realistic LLM integration in data workflows, emphasizing where current solutions fall short and how to bridge these gaps for practical implementation.

Unveiling Challenges for LLMs in Enterprise Data Engineering