Full stack Engineer - AI
Ebury helps ambitious businesses unlock global growth, and we take the same approach with our people. We encourage innovation and movement, collaboration and problem-solving, and foster an environment where everyone can feel they belong, are valued, supported and empowered to succeed. If you’re a collaborator who wants to help transform how businesses operate globally, get in touch - we’d love to discuss how Ebury can accelerate your career so you can shape the future.Senior Full-Stack Engineer — Flow (AI Agent Platform) Location: London Victoria Office Work Pattern: Hybrid (4 days office, 1 day remote) Team: Flow (Core AI Infrastructure) About the Role Flow is our internal AI agent platform. This isn't a basic chatbot bolted onto a help center; it is core agentic infrastructure that reaches across our data warehouse, services, and central systems to execute complex work on behalf of the business. Flow powers both internal and client-facing workflows, delivering true automation and replacing traditional ticketing queues with instant, intelligent actions. We are looking for a Senior Full-Stack Engineer to help us transition Flow from a fast-moving internal product to durable, productionized platform infrastructure. You will own complex features end-to-end—including agent workflows, reasoning UIs, and cross-system integrations—while raising the engineering bar and shaping the technical direction of a small, autonomous team. This role sits at the intersection of agentic AI and live operational workflows within a regulated fintech environment. The surface area is the entire company, and your impact will be visible within weeks. What You'll Do Build Action-Oriented Agent Workflows: Design multi-step orchestration, tool use, and retrieval grounded in our data stores using the Strands Agents SDK. Architect End-to-End Features: Own development across a React/TypeScript frontend and Python/FastAPI backend, balancing rapid iteration with platform maintainability. Own the Reasoning Layer: Extend our Chain-of-Thought UI to ensure explainability. In a regulated environment, showing why an agent made a decision is a core feature, not a nicety. Make Agents Measurably Reliable: Build and extend our evaluation harness (Ragas) and observability platform (Langfuse, with session continuity) so we ship changes based on evidence, not vibes. Define Evaluation Strategies: Build custom evaluators, datasets, benchmarks, and automated regression suites to catch quality regressions before they hit production. Integrate Systems & Clouds: Connect internal APIs, data warehouses, email, and cross-cloud GCP/AWS services with robust error handling and distributed tracing. Industrialize Prototypes: Drive the discover → industrialize → productionize lifecycle, turning promising AI prototypes into hardened, daily-deployed services on CD pipelines. Mentor & Lead: Uplift mid-level engineers through code reviews, pair programming, and establishing scalable engineering patter...