AI Workflow Architect

A systems-level thinker who transforms scattered AI experiments into production-grade orchestration pipelines that survive real users, messy data, and organizational politics — with the infrastructure discipline to know that week-one architecture decisions determine whether a system is maintainable in month six or becomes untouchable debt.

- The RENDER methodology — 6-pillar framework from requirements decomposition to evolution governance

- Multi-step LLM chain architecture with state passing, retry logic, and conditional branching patterns

- Agent loop design with bounded iteration limits, escape hatches, and tool-use governance

- Failure mode taxonomy distinguishing model degradation, data drift, tool errors, context overflow

- Model routing strategies matching task complexity to cost-appropriate tiers with caching layers

- RAG pipeline design with chunking, embedding selection, retrieval scoring, and re-ranking

- Hybrid human-AI handoff design with queue management, SLA routing, confidence-threshold escalation

- Production readiness criteria: deterministic boundaries, cost ceilings, graceful degradation, version control

- Structured logging and instrumentation for non-deterministic systems with trace IDs and output fingerprinting

Drop into Claude, ChatGPT, Cursor, or any AI tool. Bring your real AI workflow problem — a multi-step chain that's failing silently, an agent loop that's runaway-executing, a cost model you can't see, a RAG pipeline leaking context. It thinks like an ops engineer who's wired LLM chains through production and rebuilt architectures under pressure.

Bundles or prompts related to AI systems architecture and orchestration engineering.