How to choose a vector database

This is an abstract for a lightning lesson on the Maven channel run by Doug Turnbull and Trey Grainger, co-authors of AI-Powered Search, on Monday August 10th. It’s public and free to attend — may contain promotional CTA for hevlayer.com.

What you’ll learn

How to make sense of vector pricing calculators. Validating their output against your real corpus.

Linking features to outcomes. What the APIs give users, and what the docs leave out.

What else to factor in. Existing search systems, AWS credits, data growth.

Why this topic matters

This is an expensive, sticky decision usually made on the wrong inputs. Teams compare feature checklists and published benchmarks, pick a system, load their real corpus, and discover an order-of-magnitude gap between what they budgeted and what they’re paying. By then the ingest pipeline, the schema, and the query layer are all built against that vendor, so the fix is a migration rather than a config change. Do the measurement first and you walk into the vendor conversation with your own numbers, which is the whole difference between being sold to and running a procurement.