AI Engineer - Internship

Overview

Within our Client AI & Analytics team (CAIA), we have built our own multi‑agent AI platform from scratch. The platform is used by commercial colleagues such as private bankers, relationship advisors and financial advisors to support a wide range of use cases, including client summaries, end‑of‑year reviews, meeting notes, investment proposal reviews and other daily advisory workflows.

The platform combines frontier AI models, agentic workflows and a growing set of capabilities across text, voice, memory and personalisation. We are moving beyond one‑size‑fits‑all AI towards a platform that can better understand how colleagues work, communicate and serve their clients.

This internship focuses on GEPA‑based self‑optimisation. GEPA, or Genetic‑Pareto, is a reflective optimisation framework that uses evaluation signals, traces and natural‑language reflection to evolve prompts and other textual components of AI systems. It can be applied to prompts, instructions, agent configurations and broader compound AI systems.

The role gives you the freedom to shape your own direction. You may investigate how GEPA can optimise existing agents based on live interaction data, improve user‑level memory through implicit behavioural signals, or tune smaller, more cost‑efficient models to approach the quality of frontier models. The common thread is clear: using GEPA to explore how our platform can learn from real usage and improve safely, measurably and continuously.

As an AI Engineer Intern at Van Lanschot Kempen, you will research, design and prototype GEPA‑based optimisation loops for our CAIA platform, exploring how interaction traces, feedback signals and evaluation metrics can help improve agents, prompts, memory and model selection safely, measurably and continuously.

• Use GEPA to optimise agent instructions, tool definitions and workflow configurations.
• Explore GEPA‑based personalisation of user memories and behavioural preferences.
• Investigate GEPA tuning of smaller models to reduce costs without losing quality.
• Build evaluation loops for measuring agent performance, reliability and user satisfaction.
• Prototype self‑optimising mechanisms for selected CAIA platform use cases.
• Analyse live interaction data to identify patterns, friction points and improvement opportunities.
• Speak with commercial colleagues to understand how they use AI in their daily work.

• Currently completing or have a degree in Computer Science, Artificial Intelligence, Data Science or a related technical field.
• Experience building applications or prototypes with LLM APIs such as OpenAI, Anthropic or Gemini.
• Familiarity with LLM concepts such as prompts, tool calling, structured outputs and evaluation.
• Basic front‑end experience with Angular, React or Vue.
• Backend development experience in Python or C#.
• Good mastery of English, both oral and written.

Curious, self‑motivated and resourceful. You should be able to be given a goal, think critically about how to approach it and independently find the best path forward. You enjoy exploring unfamiliar topics, testing different solutions, asking the right questions and turning ambiguity into progress. Combine technical curiosity with a practical mindset and enjoy building things that real users can work with.