Senior AI Engineer

This is a contract role - Remote - Upto 30,000 AED/ monthSenior LLM / AI Engineer – Agentic AI & Platform EngineeringRole SummaryWe are building an AI-powered enterprise platform that uses agentic AI to transform unstructured domain data into structured, actionable intelligence. As a Senior LLM / AI Engineer, you will design and build the core AI orchestration layer: multi-agent workflows, retrieval-augmented generation pipelines, tool-calling infrastructure, and the deterministic reasoning systems that make LLM outputs reliable enough for production use.This is not a research role. You will build production application services that real users depend on. You need to be equally strong in LLM engineering (prompting, structured outputs, evaluation) and software engineering (APIs, databases, deployment). We are looking for someone who can architect an agentic system from scratch, implement it in FastAPI, evaluate it rigorously, and operate it in production.Key ResponsibilitiesAgentic AI & Multi-Agent OrchestrationDesign and implement multi-agent systems where specialised agents collaborate on complex tasks: planning agents, execution agents, validation agents, and human-in-the-loop review stepsBuild agentic workflows using LangGraph, LangChain, or custom state machines with clear state management, conditional routing, retry logic, and graceful failure handlingImplement tool-calling infrastructure: define tool schemas, manage tool registries, handle tool execution with timeouts and error recoveryDesign context management strategies for long-running agent sessions: conversation memory, working memory, context window optimisation, token budget management, and multi-turn state trackingRAG & Knowledge RetrievalBuild and optimise retrieval-augmented generation (RAG) pipelines: document chunking strategies, embedding model selection, vector store management (FAISS, Qdrant, pgvector, Pinecone), and hybrid searchImplement advanced RAG patterns: multi-hop reasoning, reranking, query decomposition, self-querying retrieval, and citation groundingDesign and maintain knowledge indexing pipelines that ingest, transform, and index domain-specific data at scaleDeterministic Workflows & Structured OutputDesign deterministic workflows that guarantee consistent, reliable outputsfrom LLMsImplement comprehensive input/output validation using Pydantic models, JSON Schema constraints, and structured output parsingBuild hybrid pipelines combining deterministic business logic with LLM-powered reasoningEvaluation, Testing & OptimisationBuild evaluation frameworks for LLM-powered features: automated test suites, regression benchmarks, and continuous production monitoringImplement token optimisation strategies: prompt compression, caching, response streaming, batching, and model selectionDesign A/B testing infrastructure for prompt variants, model versions, and pipeline configurationsPlatform & Application DevelopmentBuild production APIs using FastAPI: endpoint design, async handlers, WebSocket/SSE for streaming responses, and graceful degradationDesign and implement MCP (Model Context Protocol) server setups for standardised tool integrationWork with PostgreSQL, Redis, and Elasticsearch for application state, caching, and searchContainerise AI services with Docker and collaborate with DevOps on Kubernetes deploymentand scalingRequired Qualifications4+ years of software engineering experience, with at least 2 years focused on LLM-powered application developmentin productionStrong proficiency in Python; experience with FastAPI or similar async frameworks requiredDeep hands-on experience with agentic AI frameworks: LangGraph, LangChain, or custom agent architecturesProven experience building RAG systems: embedding pipelines, vector databases, retrieval strategies, and multi-hop reasoningStrong understanding of LLM fundamentals: prompting, structured output, token management, and model selection trade-offsWorking knowledge of PostgreSQL, Redis, and at least one vector databaseExperience with evaluation and testing of LLM outputs in production environmentsSolid understanding of REST APIs, WebSockets, and API lifecycle managementPreferred ExperienceExperience with MCP architecture: building MCP servers and integrating external systemsas agent toolsFamiliarity with parameter-efficient fine-tuning methods (LoRA, QLoRA)Experience building multi-tenant AI applications with organisation-level isolationLLM observability experience: tracing, token usage tracking, latency profilingFamiliarity with Elasticsearch and hybrid search(BM25 + vector)Cloud deployment experience (AWS preferred): EKS,S3, SQS, LambdaContributions to open-source AI/LLM projects or published work in NLP or agentic AI