
Talking Serverless #68 - Databases in Higher Dimensions: A Conversation with Jack Pertschuk | Talking Serverless #68
Jul 29, 2025
Jack Pertschuk, Principal Engineer of Algorithms and Platforms at Pinecone, dives into the world of vector databases and their pivotal role in AI applications. He shares insights on architecting scalable LLM-centric systems and the nuances of handling high-dimensional data. Jack discusses innovative use cases, from biotechnology to AI image classification, emphasizing the significance of metrics and experimentation. He also highlights Pinecone's ease of integration in serverless architecture, transforming how engineers work with data.
AI Snips
Chapters
Transcript
Episode notes
Pinecone Coined 'Vector Database'
- Pinecone's team claims they coined the term "vector database" about four years ago.
- Jack Pertschuk describes early adoption by big tech and Pinecone's founding role.
Vector Databases Vs SQL
- Vector databases use embeddings instead of SQL queries to find related content.
- Jack Pertschuk explains embeddings act as high-dimensional model state for search.
Retrieval Flow For LLM Chatbots
- User text becomes an embedding that triggers a vector search to fetch relevant snippets.
- Pinecone returns the raw text alongside vectors so the LLM can use it as context.

