Building the Production Infrastructure Your Agents Actually Need

Remember when everyone rushed to ship microservices without service discovery, circuit breakers, or tracing? AI agents are in that exact "wild west" phase right now. Everyone is building them; very few are thinking about the infrastructure underneath.

Drawing from real-world experience deploying production agents across 120+ microservices, this talk bypasses the basic tutorials and maps out the concrete stack that’s actually emerging:

• Tiered Memory & Context: Why basic RAG isn't enough, and how agents need to manage short-term, long-term, and episodic memory.

The Observability Crisis: You can't `tail -f` an agent. We'll look at tracing reasoning chains, confidence signals, and execution paths.

The Testing Gap: How to unit test non-deterministic behavior and catch regressions before your users do.

Whether you're struggling with single vs. multi-agent orchestration or just trying to keep your agent from hallucinating its budget away, this talk draws direct parallels to how we eventually fixed the microservices mess - and calls out what’s still painfully missing.

Register at weblink

Speaker: Uday Kanagala, Navan