Databricks · Databricks Generative AI Engineer Associate
Chunking
Databricks Generative AI Engineer Associate · Lesson 4 of 17 · 21:08
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the Databricks Certified Generative AI Engineer Associate exam. In this lesson: chunks that fit the embedding limit and the chat window, chunk sizes and overlap, recursive and structure-aware splitting, semantic, parent-and-child, and hierarchical chunking, and writing chunks to a Delta table with one row per chunk, a primary key, and change data feed. Includes three original practice questions.
Preview the topics
- Copperline embeds its manuals with a model whose limit is 512 tokens, but its chunks are about 1,500 tokens, and answers often miss facts from the second half of a section. What is the most likely cause?
- Copperline’s troubleshooting chunks are small and match questions well, but the answers lack the steps before and after the matched passage. Which strategy fits?
- Copperline’s chunks sit in a Delta table with an ID column and a text column, but creating a Delta Sync index on that table fails. What is most likely missing?