Talking Serverless

[Video Episode] Databases in Higher Dimensions: A Conversation with Jack Pertschuk | Talking Serverless #68

Jul 29, 2025
Jack Pertschuk, Principal Engineer at Pinecone, dives into the world of LLM-centric architectures and vector databases. He reveals what it takes to build reliable and scalable LLM applications. The discussion includes the significance of vector embeddings in enhancing data retrieval and chatbot functionality. Jack discusses Pinecone’s innovative approaches to recommendation systems using 'vibe tags.' He also highlights the evolution of serverless technologies and the integration of probabilistic methods in AI applications, providing insights for both developers and enthusiasts.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Vector Databases Enable Semantic Search

  • Vector databases search by similarities in high-dimensional vectors, not exact matches like traditional databases.
  • They excel at retrieving related content by using embeddings derived from text or other data types.
ADVICE

Store Text with Vectors

  • Store the original text alongside the vector embeddings for retrieval.
  • This avoids the infeasibility of reversing embeddings to reconstruct the original text.
ANECDOTE

Pinecone's Original Name: Hypercube

  • Pinecone was originally named Hypercube, inspired by the four-dimensional cube concept.
  • They rebranded to Pinecone to better fit their vision and branding.
Get the Snipd Podcast app to discover more snips from this episode
Get the app