Presentation: Multi-Agent Patterns from Spotify’s AI Powered Advertising Platform
Spotify runs production multi-agent systems at scale—here's what actually broke and how they fixed it.
Why it matters
A detailed practitioner case study on hardening multi-agent architectures in production: domain ownership, deterministic guardrails, tool schema optimization, and cost management. Rare real-world engineering patterns from a high-scale deployment.
The key facts
8 to knowSpotify Ads Manager runs production-grade multi-agent systems using Google ADK Java
Key architectural patterns: domain ownership models, deterministic guardrails, tracing-based evaluation
Hard-learned lessons on agent boundary definition, tool schema optimization, and cost control
Addresses monolithic agent pitfalls and multi-agent orchestration at scale
Presentation format; no new product announcement or GA release
Hard-learned lessons on agent boundary drawing, tool schema optimization, cost management
Avoids monolithic agent pitfalls through identified architectural patterns
Presented by Pratik Rasam at InfoQ (Oct 8, 2026)
Go to the source
InfoQ AI/MLinfoq.com
Publisher excerpt: Pratik Rasam discusses how Spotify Ads Manager runs production-grade multi-agent systems at scale using Google ADK Java. He shares key architectural patterns, domain ownership models, deterministic guardrails, and tracing-based evaluation strategies, detailing hard-learned lessons on drawing agent…