AI engineer is the fastest-growing engineering title in 2026, and one of the least well-defined. At most companies it means someone who builds production applications on top of large language models — RAG systems, agentic workflows, AI-powered features — rather than someone who trains models from scratch (that is still the ML engineer). The CVs getting filtered out are ones that list "prompt engineering" as a primary skill with no evidence behind it, or that show only tutorial-grade projects: ChatGPT wrappers and simple RAG demos that do not prove production readiness. The CVs that get shortlisted show two things the market is genuinely short on: rigorous LLM evaluation — systematic testing of whether the AI system actually works as intended — and production economics — having shipped AI features that operated within latency and cost constraints for real users. This guide covers the ATS keywords, structure, and bullet patterns that position an AI engineer CV for the roles that are actually in demand.

What AI Engineer Job Descriptions Require in 2026

AI engineer JDs have converged around a consistent technology stack despite the title being new. The companies hiring for this role in 2026 expect:

  • LLM API integration — OpenAI, Anthropic Claude, Google Gemini, and open-source models (Llama, Mistral via Ollama or HuggingFace). JDs name specific providers; your CV should match. Most companies use 1–2 providers and want engineers who have production experience with them, not just API familiarity.
  • RAG architecture — Retrieval-Augmented Generation is the dominant pattern for enterprise AI applications in 2026. JDs expect: document chunking and embedding strategies, vector database integration, hybrid search (semantic + keyword), and reranking. This is more complex than "call the embedding API and store in Pinecone."
  • Orchestration frameworks — LangChain and LlamaIndex remain the most common; newer alternatives like LangGraph (for agentic flows), CrewAI, and AutoGen appear in JDs at companies doing multi-agent work. Python is assumed throughout.
  • Vector databases — Pinecone, Weaviate, Chroma, Qdrant, and pgvector (PostgreSQL extension). JDs often name a specific one; understand the trade-offs between them.
  • Evaluation and observability — LangSmith, RAGAS, DeepEval, and Arize AI for LLM evaluation; monitoring for hallucination rates, faithfulness, and context utilisation. Increasingly listed as a requirement, not a bonus.
  • Production engineering — FastAPI for serving AI endpoints, streaming responses, cost monitoring and optimisation, prompt caching, model routing between cheap and expensive models based on query complexity.
  • Fine-tuning (for specialised roles) — LoRA/QLoRA for efficient fine-tuning of open-source models; PEFT libraries from Hugging Face. Appears in JDs at companies that cannot use general-purpose LLMs due to data privacy or cost.

AI engineer salaries in 2026: £55K–£95K UK; $120K–$185K US, with senior AI engineers at product companies exceeding $200K total compensation. The market is undersupplied at the senior level.

ATS Keywords for an AI Engineer CV

AI engineer is a new title, which means ATS keyword conventions are still forming. Your CV should include both the title variants and the specific technology names that JDs use.

Essential ATS terms for an AI engineer CV:

  • Title variants: AI Engineer, LLM Engineer, Generative AI Engineer, AI Developer, Prompt Engineer, AI Software Engineer, ML Engineer (for cross-discipline roles)
  • LLMs and providers: OpenAI, GPT-4, Claude, Anthropic, Gemini, Llama, Mistral, Hugging Face
  • Frameworks: LangChain, LlamaIndex, LangGraph, CrewAI, AutoGen, DSPy
  • Vector databases: Pinecone, Weaviate, Chroma, Qdrant, pgvector, FAISS
  • Techniques: RAG, retrieval-augmented generation, embeddings, semantic search, prompt engineering, fine-tuning, LoRA, QLoRA, agentic AI, multi-agent systems, tool calling, function calling
  • Evaluation: RAGAS, DeepEval, LangSmith, Arize AI, evaluation pipelines, hallucination detection
  • Infrastructure: FastAPI, Docker, Kubernetes, AWS SageMaker, GCP Vertex AI, Redis, streaming, prompt caching
  • Long-tail phrases: AI engineer resume examples, how to write an AI engineer resume, LLM engineer resume, generative AI engineer cv, AI developer skills for resume

Placement: Python and your primary LLM framework (LangChain or LlamaIndex) belong in your headline. "RAG" and "LLM" as explicit terms in your Skills section improve ATS matching significantly — these are the keyword terms recruiters search for alongside "AI engineer."

AI Engineer CV Structure and Bullets That Prove Production Capability

Section order:

  1. Headline — "AI Engineer | Python · LangChain · RAG · Claude · Pinecone" or "AI Engineer | LLM Applications · LangGraph · OpenAI · FastAPI"
  2. Skills — layered: Languages / LLM Frameworks / Vector Databases / Infrastructure / Evaluation
  3. Experience — 4–6 bullets per role; user-scale metrics and evaluation results here
  4. Projects — list 2–3 production or near-production AI applications with specifics (LLM used, architecture, user metrics); GitHub or demo links
  5. Education — CS, mathematics, or linguistics degree; certifications (DeepLearning.AI, Anthropic, OpenAI) are worth listing

One to two pages; one page for under 2 years (the title is new enough that senior candidates may only have 2–3 years specifically as AI engineers).

The three data points that make an AI engineer bullet stand out: the AI architecture used, the evaluation metric achieved, and the production impact. Three examples:

  • Built a legal document analysis RAG system using Claude with pgvector on PostgreSQL, processing 12,000 client documents with hybrid semantic-keyword search and cross-encoder reranking; achieved 94% answer faithfulness on the evaluation set using RAGAS and reduced average document review time from 3.5 hours to 22 minutes
  • Designed a multi-agent customer support pipeline using LangGraph with tool-calling for CRM lookup and ticket creation, handling 8,000 daily queries with 78% first-contact resolution without human escalation and reducing support team volume by 40%
  • Reduced LLM API spend on a document summarisation pipeline by 62% — from £18K to £6.8K monthly — by implementing Anthropic prompt caching, model routing between GPT-4 and GPT-3.5 Turbo based on query complexity, and Redis caching for duplicate queries, while maintaining response quality within 2% of baseline RAGAS faithfulness scores

AI engineer interviews in 2026 are still forming their conventions, but common patterns include: a system design question ("design a customer-facing AI assistant that can answer questions about our product catalogue"), a prompt engineering exercise (show how you would improve this prompt's output quality), and a discussion of evaluation strategy (how do you know your AI system is working). Candidates who can discuss evaluation methodologies and production cost constraints stand out sharply because most candidates cannot.

Three AI Engineer CV Mistakes That Signal Tutorial-Grade Experience

"Prompt engineering" listed as a primary skill with no evaluation evidence. Prompt engineering is real work — but it is also the skill claimed by every person who has used ChatGPT. The differentiator is whether your prompt engineering was systematic: version-controlled prompts, A/B tested against a held-out evaluation set, measured for output quality metrics (faithfulness, relevance, groundedness). Without this evidence, listing prompt engineering as a primary skill signals amateur usage, not professional engineering. Add one metric: "improved RAG answer faithfulness from 71% to 89% through systematic prompt optimisation measured using RAGAS" is conclusive.

Portfolio showing only tutorial-grade demos. Most AI engineer portfolios contain: a ChatGPT wrapper, a simple RAG over a PDF, and a chatbot built with LangChain following a YouTube tutorial. Hiring managers have seen hundreds of these. Production evidence — an AI system that handled real user traffic, within latency and cost constraints, with monitoring and evaluation in place — is what separates the candidates who get offers. Even a side project that served 200 users with real error tracking and evaluation logs is more convincing than a polished demo that was never used.

No cost or latency metrics. LLM API calls are expensive and slow relative to traditional software. Production AI engineers must operate within cost and latency budgets — this is a core part of the role that tutorial-grade engineers never encounter. A CV with no mention of API cost, latency SLAs, or cost optimisation strategies reads as someone who has not yet shipped AI features to real users under production constraints. Include one metric: "reduced average query cost from $0.04 to $0.009 through model routing and prompt caching" signals economic awareness that distinguishes production engineers from experimenters.

If you are applying to AI engineer or LLM engineer roles and want your CV rebuilt around the specific RAG, LangChain, and evaluation requirements in a target job description, Resumegpt generates your AI engineer CV from your work history in under 60 seconds — ATS-optimised, production signals included, and exported as a PDF ready to submit.