that check outputs against the required format and quality, and use the results to improve the solution. Partner with engineering and AI specialists on LLM orchestration, tool calling, RAG, context management, prompt iteration, model selection, and cost-performance trade-offs. Design human-in-the-loop controls for cases where … iteration, ideally across conversational AI, voice agents and telephony, workflow automation, decision support, diagnostics, or intelligent self-service. Strong working knowledge of multimodal models, LLM capabilities and limitations, prompt engineering, tool calling, embeddings, RAG, and AI observability. Fluency with evals: defining what good output looks like, building eval sets ...