Vector Search, From the Whiteboard to the Cloud

Wed, 01 Jul 2026 00:00:00 +0000

🎬 Also available as a blog walkthrough video: a narrated screencast of this post.

TL;DR: Vector search is easy to picture on a whiteboard: turn text into a list of numbers, then find the stored items whose numbers point in a similar direction. That mental model is correct and it is enough to reason about retrieval. Production is where it gets subtle: chunking, recall versus latency, embedding drift, and metadata filtering all decide whether retrieval helps or quietly hurts. On AWS, Amazon Bedrock Knowledge Bases runs that whole loop as a managed service (ingestion, chunking, embedding, and vector storage) across vector stores like OpenSearch Serverless and Aurora PostgreSQL. This is Part 1 of a three-part series, and the first post in a new “Whiteboard to Cloud” format that pairs an academic explainer with the AWS practice.

A few months ago I argued that RAG is still needed, even with million-token context windows [1]. That post was about when to retrieve. This one goes one level down: what is the retrieval actually doing?

schristoph.online

Vector Search, From the Whiteboard to the Cloud