(Senior) Back-end engineer Conversational AI

About us & the role

We believe customer interactions are evolving toward AI agents that actively support personalized financial choices. These agents do not operate in isolation: they collaborate with other systems, services, and agents to reason, act, and deliver outcomes at the right moment. Screens and integrations remain essential, but their role shifts, from static interaction points to dynamic execution and orchestration layers that expose insights, actions, and context exactly when needed.

That's why we are looking for a Senior Backend Engineer to strengthen our Conversational AI teams and accelerate the delivery of high-impact AI-powered features for our AI Agent Robin. This includes leading backend initiatives focused on integrating AI agents with existing services, enabling secure context sharing, and supporting agent-to-agent (A2A) interactions and Model Context Protocol (MCP)-style patterns to create richer, more capable customer conversations.

Your main focus will be on

• Designing and building backend services that power key conversational features for our AI agent Robin (e.g., financial insights, mortgage scenarios, transaction explanations, and document generation), with a strong focus on AI-agent-driven workflows.
• Integrating Large Language Models (LLMs) and AI agents with internal systems using well-defined APIs, event-driven patterns, and secure context propagation.
• Enabling agent orchestration and A2A-style interactions, allowing Robin to collaborate with other agents or services (e.g., planning, reasoning, and execution agents) using MCP-inspired patterns.
• Developing reusable backend components and services that can be leveraged across multiple conversational agents, ensuring the platform is scalable, observable, and future-proof.
• Working closely with front-end, integration, and platform engineers to ensure seamless data, intent, and context flow between the chat interface, AI agents, and core banking systems.
• Collaborating with business analysts, designers, data scientists, and fellow engineers to shape conversational capabilities, define backend best practices, and decide how responsibilities are split between agents, services, and traditional application logic.

• Senior-level experience in backend software engineering, je hebt 5+ years of professional experience building and operating production-grade services.
• Hands-on experience integrating AI/LLM-based components into backend systems, including prompt orchestration, tool/function calling, and context management.
• Experience with distributed systems and API design, and familiarity with event-driven or message-based architectures.
• A strong sense of ownership je bent proactive, quality-driven, and comfortable making architectural trade-offs in complex environments.
• A customer- and outcome-focused mindset, je understandt that backend design directly impacts customer trust, explainability, and experience.
• Backend development using Python and Java.
• Designing and implementing REST APIs, MCP and A2A integrations for AI-agents and applications.
• Experience with AI-agent architectures, including tool invocation, context persistence, conversation state, and guardrails.
• Event-driven architectures, messaging platforms, and distributed system patterns.
• Security, privacy, and compliance considerations when exposing data to AI agents.
• Observability (logging, metrics, tracing) and reliability engineering in AI-augmented systems.

Wensen

• Ideally, je have previously worked at Rabobank and are familiar with our backend and integration landscape.
• Experience with low-code conversational platforms like copilot studio is a pre

Youll work in a hybrid setup, ideally spending one day a week onsite in Utrecht, collaborating closely with teams to implement new conversational and agent-integration concepts. On Mondays, our department works from the office. For the rest of the week, we coordinate among ourselves whether we work from home or the office. The Digital & Customer Interaction Tribe aims to deliver an excellent customer experience, regardless of how the customer contacts Rabobank. Core values of our area include fun, collaboration, proactivity, and problem-solving.

Facts & Figures

• 8 million customers

• 43,822 Rabobank colleagues worldwide

Optie tot verlenging: Niets over bekend.

Hybride: Youll work in a hybrid setup, ideally spending one day a week onsite in Utrecht, collaborating closely with teams to implement new conversational and agent-integration concepts. On Mondays, our department works from the office. For the rest of the week, we coordinate among ourselves whether we work from home or the office.

Fee: De administratieve partner rekent een aanvullende fee van € 2,50 boven op het uurtarief.

Overig algemeen:

Apply

To ensure that we assess all candidates equally for suitability for this position, we do not require a cover letter. Instead, we ask all candidates the same three questions to see how your profile matches the position. Pay close attention to answering these questions and explaining your answers; this, together with your CV, is the only information we will use to evaluate je in this round. If this is missing from your application, we will not be able to consider your application.

Use a maximum of 200 words per question and provide no more than two examples per question.

Questions:

• How would je design a backend architecture that enables an AI agent to safely interact with multiple internal services while maintaining context, auditability, and security? What patterns would je apply to keep the system scalable and loosely coupled?

• Imagine Robin needs to collaborate with other specialized agents (e.g., planning or risk-assessment agents). How would je enable agent-to-agent (A2A) communication, and how would je manage shared context and responsibilities between them?

• A new AI-driven conversational feature causes increased latency and unpredictable load on downstream systems. How would je diagnose the issue? What measures would je take to improve performance and ensure reliability without limiting future experimentation?