Artificial Intelligence Engineer

Agentic AI Platform DeveloperLocation: EuropeStart: ASAPDuration: 6 monthsContract type: B2B/ FreelanceAbout the role:Remobi is looking for an Agentic AI Platform Developer to join the team building and operating an Enterprise Agentic AI Platform for a Finance organisation. This platform gives Finance professionals AI-powered specialist agents that search enterprise data, navigate policies, automate compliance checks, and manage enterprise architecture, business process, and approval workflows — all through natural language conversation.The platform serves real users today: controllers who ask to see maverick spend by business unit, procurement managers who need policy guidance, and investment planners who track capital proposals. Every feature built reaches real users the same week it's deployed.You'll join the team that builds and operates this platform — from the agent intelligence layer to the cloud infrastructure that runs it.What You'll Actually Do:Build and maintain AI agents that connect to enterprise systems. Each agent has deep domain knowledge (procurement, investment, FX, policy, enterprise architecture) and uses specialised tools to query data, create artifacts, and take action on behalf of usersDevelop and operate MCP/CLI servers — the protocol layer connecting agents to enterprise data sources, knowledge bases, and business systems; this is the backbone of the platformShip to cloud — containerise, deploy, and monitor on AWS; write infrastructure-as-code, manage CI/CD pipelines, and ensure the platform is secure, observable, and reliable for multi-user operationDesign agent experiences — write the system prompts and tool-calling workflows that make agents useful, not just clever, optimising for accuracy, speed, cost, and user trustWork across the full stack — this is platform engineering, not feature development. One week you're writing a Terraform module for a new environment; the next you're tuning a RAG pipeline for better search relevance; the week after you're debugging why an agent called the same tool 25 times in a loopRequired Skills and Experience:Backend development in Python or TypeScript — the platform's MCP servers, ETL pipelines, and agent orchestration are primarily Python, with workflow engines and some tooling in TypeScriptDocker and containerisation — the platform runs as containers, locally via Docker Compose (14 containers) and in the cloud via ECS Fargate; comfortable writing Dockerfiles, debugging container issues, and understanding multi-stage builds, health checks, and container networkingCloud infrastructure on AWS — ECS Fargate, S3, RDS, Cognito, Secrets Manager, ALB; experience deploying and operating services on AWS, with Infrastructure-as-Code via Terraform (no click-ops)Git, GitLab, and CI/CD — comfortable with branching strategies, pipeline configuration, and automated deployments using a self-hosted GitLab instanceSQL and working with data — able to write SQL, understand data modelling, and reason about query performance across federated query engines, local analytics databases, and cloud data warehousesUnderstanding of LLMs — you don't need to train models, but you need to understand how large language models generate responses, how token limits affect design decisions, and how to evaluate whether an agent's output is good enough; daily work with cloud LLM APIsDesirable Skills:Experience with the Model Context Protocol (MCP), or willingness to learn it fast — MCP is the foundational protocol of this platform, with every agent interacting with enterprise systems through MCP servers exposing tools via stdio or SSE transportExperience building RAG pipelines — the platform's knowledge layer uses vector search with hybrid retrieval (keyword + vector + cross-encoder reranker) over 300+ documentsExperience with agent system prompt engineering — each agent has a 200-600 line system prompt defining its behaviour, query patterns, tool usage, governance rules, and domain knowledgeAuthentication and identity experience — the platform uses cloud identity providers with SAML federation for SSO, with access control via enterprise identity governance roles mapped to per-agent permissionsWorkflow and orchestration design — experience with state machines, event-driven architectures, or workflow engines, relevant to the platform's approval-gated workflowsNice to Have:AI observability tools experience — tracing LLM calls, tool invocations, and agent decisions for quality scoring, cost tracking, and debuggingQuery federation or analytical database experienceFinance domain knowledge — cost centres, GL accounts, procurement processes, investment proposals, or FX operationsEnterprise system integration experience — ERP systems, enterprise architecture tools, dashboard platforms, or similarAI governance awareness — NIST AI RMF, EU AI Act, GDPRFrontend skills — React, CSS customisation, or building conversational UIs, relevant to the platform's custom chat interface with interactive charts, process diagrams, and visual artifactsTech stack highlights: Python, TypeScript, Docker, AWS (ECS Fargate, Cognito, S3, RDS), Terraform, GitLab CI/CD, federated query engines, embedded analytics databases, vector search, AI observability