article Sep 06, 2025Lightweight Re-Ranking in Qdrant Using Logistic RegressionI’ve worked a little more with Qdrant’s hybrid queries and noticed that they are useful beyond what I described in my last article. When building a recommendation system, we usually start with vector search to retrieve promising candidates — for example, the top-M videos for a user based on their past behavior in a shared embedding space.[... 409 words]pythondata-sciencevector-db
article May 23, 2025Combining Vector Search with Business Logic in QdrantI’ve been using Qdrant for vector similarity search in a recommendation system, and one challenge that keeps coming up is how to incorporate additional information beyond just the vectors themselves.[... 353 words]pythonsqldata-sciencevector-db
article Mar 02, 2025A Lightweight Vector DB with DuckDB and Cloud RunWhile working on a project, I recently needed a lightweight, serverless vector database solution. The vectors needed to be stored alongside additional metadata, and the database itself would only be updated daily—i.e., read-only for most users.[... 518 words]data-engineeringcloudgcppythonduckdbvector-db