Code-driven agent orchestration
Code: ENG-003 | Track: technical | Duration: 0.5 day
Designing reliable, stateless, multi-agent orchestrated architectures: linear sequences, concurrent execution and smart routing.
Context
Teams moving from a single agent to orchestrated, reproducible multi-agent architectures.
Who it's for
Engineering teams
Prerequisites
- Working Python or TypeScript environment with API access; starter repo shared in advance
Format
- Remote
Agenda
- Workflows vs agents - choosing the right architecture - when simple, stateless workflows beat looping agents, and the core principle of using the simplest architecture that solves the problem.
- Core orchestration patterns - prompt chaining, routing and tool use, with a guided code-along through a routed request.
- Building prompt-chained workflows - hands-on lab decomposing tasks into fixed-step pipelines with predictable costs and error rates.
- Failure modes and error mitigation - error compounding, token scaling, and why retry plus fallback beats complex self-correction.
- Observability and governance - traces, logs and cost tracking per workflow, step and model call.
- Cost optimisation and production readiness - model selection, caching and batching, closed out with a pre-production checklist.
Session details
- Total time: 0.5 day (4 hours including two short breaks)
- Participants: Up to 15
- Delivery: Remote or in person (UK; other locations by arrangement)
- What you leave with: A working three-stage chain with retry, fallback and tracing, plus a pre-production checklist