You've built agentic systems — tool use, orchestration, retry/repair loops, context management — whether on raw provider APIs or frameworks like LangGraph, PydanticAI, Google ADK, Pi-mono, or Vercel AI SDK. Designing model strategy: routing, benchmarking, reliability, and cost/latency optimization Creating debuggable agent systems with tracing, metrics, alerts, and observability Researching novel approaches and shipping them to production.