We're looking for a Senior AI Engineer with deep agentic and RAG experience to join our client, a boutique IT consultancy delivering AI and modernization work for enterprise customers across the UK and Europe.Fully remote, B2B, project-based engagement months with a rate of $45-55 per hour depending on experience. Strong delivery on the first project opens the door to follow-on work.The roleYou'll join a small delivery team as the engineer who owns how the AI actually behaves. An architect is already on the project, so this is not an architecture seat. You build the agent workflows, tune the retrieval, prove the quality and get it running in production, and you do it while sitting in front of the client.This is an engineering role with client exposure, not a strategy or advisory role. You write code daily and you also explain, in plain language, why the system answered the way it did. The core technical challenge is reasoning over many heterogeneous data sources: how to combine them, how the reasoning loop should behave, and when to stop and return a concrete answer.What you'll doAgentic systems and orchestrationDesign and implement agent workflows in Haystack or LangChain/LangGraph: planning, tool use, multi-step reasoning, routing between sourcesDefine stopping and exit criteria for reasoning loops: confidence thresholds, iteration and token budgets, fallback paths, human-in-the-loop pointsDesign tool and function calling, enforce structured outputs, manage state across multi-step runsRetrieval and dataBuild retrieval across multiple sources: chunking, embeddings, hybrid search, reranking, graph-based retrievalMerge, deduplicate and reconcile results when sources disagree, with traceable citationsWork with vector databases and graph databases as first-class parts of the pipelineApply vision-language models where the input is visual: scanned documents, diagrams, imagesQuality and productionBuild evaluation for retrieval and answer quality, with regression checks when prompts or models changeInstrument the system: tracing of agent steps, latency, token usage and cost per requestPackage services with Docker and deploy to Azure, working with existing CI/CD pipelinesHandle the boring production parts properly: retries, timeouts, guardrails, error handlingClient workRun technical sessions with the client and answer hard questions in real timeTranslate business requirements into system behaviour and back againFlag risks early: hallucination, data access limits, cost blow-ups, latencyMust have6+ years in software engineering, expert-level Python2+ years building LLM-based systems that reached production with real usersProduction experience with Haystack or LangChain/LangGraphHands-on RAG and agentic systems, including loop control and multi-source retrievalExperience with at least one vector databaseEvaluation and tracing practice: you can show how you proved a change made the system betterDocker and CI/CD fundamentalsDirect client-facing experience: workshops, requirement discussions, presenting trade-offsFluent English, C1 minimumNice to haveMicrosoft Azure: App Services, Container Instances, networkingGraph databases (Neo4j or similar), GraphRAG or knowledge-graph-backed retrievalVLM and multimodal RAG experience (significant plus)MCP or A2APrior consultancy, agency or contracting backgroundDetailsB2B contract via PowerMind Contractors100% remote, CET and UK working hours2-3 month engagement, extension and follow-on projects possibleYou'll be working as part of our client's delivery team, under their brand, with direct contact with the end customerBackground check and hardware security key (YubiKey) required before start$45-55 / hour Show more Show less