Agentic Workflow Designer

A systems architect who engineers resilient multi-agent orchestrations that reason, adapt, and recover without human babysitting — balancing autonomy against the 80% of production work that is error handling, observability, and graceful degradation.

- The WEAVE methodology — 5-pillar framework from task analysis to production monitoring

- Multi-agent topology patterns: hub-and-spoke, pipeline, debate, hierarchical supervisor with selection criteria

- Failure mode inventory templates covering malformed outputs, tool timeouts, context overflow, infinite loops

- Retry and fallback strategies with exponential backoff, model-tier degradation, circuit breakers

- Prompt engineering for tool-calling agents with explicit instruction hierarchies and output format contracts

- Trace instrumentation and evaluation framework design with LLM-as-judge configurations

- State management and checkpoint-resumable execution for long-running workflows

- Technology stack guidance: LangGraph, CrewAI, Temporal, LangSmith, Guardrails AI, Pydantic validation

- Adversarial input testing and regression detection playbooks

Drop into Claude, ChatGPT, Cursor, or any AI tool. Bring your real agentic workflow problem — a multi-step automation that fails silently at scale, a tool-calling agent that hallucinates function names, context window overflow in long-running pipelines. It thinks like a distributed systems engineer who's shipped agent systems through customer support triage, document extraction, and code review loops.

Bundles or prompts related to AI implementation, agent architecture, and production LLM orchestration.