Google, IBM & Meta Certificates – 40% Off
One plan covers every Professional Certificate on Coursera.
Unlock All Certificates
Learn how to revolutionize AI infrastructure through asynchronous architecture in this 34-minute conference talk from the MLOps.community Agents in Production Virtual Conference. Discover how Delivery Hero reimagined GenAI infrastructure by moving beyond traditional `POST /predict` serving patterns to implement async message queues, actor-model microservices, and zero-to-infinity autoscaling without orchestrators. Explore the paradigm shift of treating every AI step as an independent async actor called "asyas" - from data ingestion and prompt construction to smart model routing, pre-processing, analysis, and backend logic. Understand how dozens of specialized actors can coexist on the same GPU cluster, each scaling from zero to required capacity, resulting in dramatically lower GPU costs, true composability, and maintainable systems. Follow the evolution from DAGs to distributed stateless async actors and see real-world production demonstrations of this architecture. Get introduced to the open-source `Asya` framework and learn about bridging async pipelines with synchronous MCP servers for real-time responses when needed. Gain insights from a Senior ML Engineer with 7+ years of experience building ML platforms and scaling global image auto-enhancement services on massive self-hosted Kubernetes infrastructure.
