Finding the perfect moment: How Formula 1® uses AWS AI to search archives in seconds

This article describes how Formula 1 uses AWS AI to enable rapid searching of its 12-petabyte video archive through a natural language query system.

• Amazon Nova interprets plain language requests and translates them into precise archive searches without replacing the existing search engine
• Entity extraction validates user queries against a catalog of F1-specific tags (drivers, teams, circuits, race conditions)
• Fact resolution handles complex queries requiring database lookups, such as identifying pole position winners
• Three-layer caching strategy reduces model costs by 88% through prompt caching and query result reuse
• Safety controls constrain the model to a closed vocabulary, prevent SQL generation, and ensure graceful fallback to plain text search
• Timeline explorer displays results chronologically, helping editors navigate decades of F1 history

The approach demonstrates how to augment existing search infrastructure with generative AI in a narrow, well-guarded role grounded in trusted data, enabling production teams to find archive footage in minutes instead of hours.