Making Vector Search work with Complex Data at ELC Annual 2024

Daniel Svonava, CEO of Superlinked, presented at the Engineering Leadership Conference (ELC) Annual 2024, a premier gathering for engineering leaders on August 27, 2024, in San Francisco. This prestigious event featured speakers from top tech companies including LinkedIn, Figma, Meta, Twilio, and Google.

Svonava spoke alongside representatives from Contextual and LanceDB. His talk focused on the challenges of vector search on structured data, highlighting a critical issue in the field:

• He addressed the problems with current approaches to vector search on structured data.
• Svonava pointed out that stringifying and embedding data using LLMs and language encoders often leads to unpredictable results.
• The presentation demonstrated how to handle complex queries using custom data and query embeddings, showcasing Superlinked’s solution to these challenges.

This talk provided valuable insights into improving vector search accuracy and reliability, especially when dealing with structured data alongside unstructured data - a crucial aspect for many engineering applications.