Encoding Your Domain Expert: The Context Layer Behind Spotify's Data Assistant
Spotify's data agent, Vedder, answers natural-language questions against 70,000+ datasets by relying on "clusters" of curated context (datasets, vetted question-SQL pairs, and docs) owned by domain experts rather than raw schema access; curators accepted only 12.5% of AI-proposed query-history examples, underscoring that human-reviewed context, not automation, is what makes the agent trustworthy.
More from Engineering Atspotify
Our Multi-Agent Architecture for Smarter Advertising
Spotify developed Ads AI, a multi-agent system using Google's Agent Development Kit and Vertex AI with specialized agents working in parallel to generate optimized media plans, reducing planning time from 15–30 minutes to seconds.
1,500+ PRs Later: Spotify's Journey with Our Background Coding Agent
Spotify built an internal CLI leveraging LLMs to automate complex code migrations and maintenance tasks across their Fleet Management platform, generating over 1,500 merged pull requests with 60–90% time savings compared to manual implementation.
