AI Engineer Job Ads Prioritize Evals and Operations Over Framework Knowledge

An analysis of 1,135 AI agent-specialist job listings from March–May 2026 found that over half mentioned no specific framework, and postings that did name frameworks carried median salaries roughly $20,000 below the segment's overall median of $213,000. When frameworks were cited, LangChain led with 392 mentions, followed by LangGraph and LlamaIndex, but rarely appeared alone — averaging 2.31 frameworks per listing. Senior job postings consistently emphasized evaluation design, observability, and operational metrics like latency and cost-per-outcome rather than tool proficiency. A late-2025 survey of 1,340 practitioners found 57% do not fine-tune models at all, with most relying on base models, prompting, and retrieval-augmented generation refined through evals. Model Context Protocol (MCP) stood out as the one cross-cutting requirement, appearing in 17–24% of postings across all five major frameworks.