Thesis intern trustworthy & AI

Company Description

Are you curious about making AI agents transparent, reliable, and compliant? Join our Trustworthy AI team and write your thesis on real challenges. At Deloitte. 

Job Description

• write your master's thesis on one of three research topics: human-in-the-loop ontology engineering, fine-tuning of domain-specific LLMs, or evaluation of AI agent behaviour in finance;

• build a working prototype in Python, such as an AI agent combined with an ontology and a reasoner, a fine-tuned language model, or a scenario-based evaluation framework;

• design experiments, define metrics and compare your approach with a baseline; 

• work with domain experts to translate informal requirements and regulations into precise, testable specifications;

• read and critically assess scientific literature and relate it to practical use cases for clients; 

• present your findings and lessons learned to your supervisor and the wider Trustworthy AI team.

Qualifications

Naturally you are keen to discover and learn new things each day, and you are curious about how AI can be used responsibly. But most importantly, be the true you. The one and only you. With your personal strengths, view of the world, and unique personality. You need the following qualifications for the position of Thesis intern: 

• you are in the final stage of your master's degree in Artificial Intelligence, Computer Science, Data Science, Econometrics, Mathematics or a comparable science master's;

• you have good Python skills and a basic understanding of machine learning and large language models;

• you are able to read scientific papers in English and to design experiments and critically assess the results; 

• you are available at least 3 days a week for at least 5 months and live in the Netherlands.

Additional Information

AI agents are increasingly used in sensitive settings such as finance and regulated industries. Organisations need to be sure that these systems are accurate, explainable and compliant. As a thesis intern in our Trustworthy AI team, you will help make that possible. 

You choose one of three research topics. Each topic is a separate thesis position: 

1. Human-in-the-loop ontology engineering with agentic AI and automated reasoning. Ontologies capture domain knowledge in a formal way. Building them from regulations is slow and demanding. You will develop a prototype in which an AI agent proposes ontology concepts and a reasoner checks them. A domain expert reviews and refines the results. The aim is a transparent and reproducible way of working, not fully autonomous generation. Experience with RDF, OWL, SPARQL or tools such as Protégé is a plus. 

2. Supervised fine-tuning for domain-specific LLMs. You will investigate whether supervised fine-tuning improves how an LLM answers domain questions, uses retrieved documents and selects tools. You will compare the fine-tuned model with its base model on answer quality, factual grounding, ontology consistency and tool-use accuracy. Experience with PyTorch, Hugging Face or LoRA is a plus. 

3. A framework for evaluating AI agent behaviour in finance. You will design simulated financial scenarios with competing objectives and regulatory requirements. You will then use them to compare agents built on different AI models, identify compliance-related failure patterns and explore ways to improve adherence. Experience with LLM APIs, scenario-based testing or pandas is a plus. 

The internship focuses on writing your master's thesis. You will be closely mentored by a thesis supervisor at Deloitte. You will also be part of the Financial Risk talent pool. Besides your thesis, you can work on projects, proposals or research for our clients. 

Please also tell us which of the three topics you prefer.