Agentic AI Engineer

We’re a fast-growing AI scale-up working with social data to understand and predict human behaviour at scale.We’re building agentic AI systems that combine behavioural science with modern AI to answer complex questions about how people and groups behave in the real world.Following our Series A, we’re looking for an Agentic AI Engineer to join our team in the Netherlands.The role has two core parts.First, you’ll work alongside our behavioural science team to build AI systems that solve specific client problems.How might a population respond to an unfolding event? How could the behaviour of a group change when its environment changes? How can large volumes of social data be turned into useful behavioural intelligence?You’ll translate behavioural science into agentic workflows that can reason across data, test hypotheses and produce useful outputs for real users.Second, You will join us in building our own foundational AI capabilities that can be reused across multiple clients.Instead of solving every new problem from scratch, you’ll identify what can be generalised and turn it into robust capabilities within our core platform.That means:Building agentic systems around behavioural science and real-world problemsWorking closely with scientists to translate research and models into AI capabilitiesDesigning multi-step reasoning and agent workflowsBuilding reusable foundational AI components across client deploymentsDeveloping tool-use, retrieval, context and evaluation capabilitiesTaking experiments from rapid prototypes into reliable production systemsWe’re particularly interested in engineers who have built real LLM or agentic AI systems and enjoy working directly with domain experts rather than receiving perfectly defined specifications.You’ll be working at the intersection of AI engineering, behavioural science and real-world decision-making.What matters is engineering depth, curiosity, comfort with ambiguity and an AI-native way of working.If building AI that turns behavioural science into real-world capabilities sounds like your kind of problem, please apply here.