PhD Position: AI Alignment and Value Assessment for Open Models & AI Systems in Law Enforcement Delft University of Technology in Netherlands

This PhD position at Delft University of Technology (TU Delft) focuses on advancing the responsible use of AI within the Netherlands Police force by developing, operationalizing, and embedding complex human and legal values into AI alignment pipelines for open-weights and foundation models. The project aims to move beyond static benchmarks, creating dynamic, automated validation and red-teaming frameworks tailored for high-risk deployments, such as law enforcement. The impact of this research is to enhance police trustworthiness and ensure compliance with the EU AI Act by engineering measurably aligned AI systems.

As a PhD student, you will investigate standards and values surrounding AI usage, especially for publicly available models. This involves defining criteria for these models, including but not limited to bias, fairness, reliability, and trustworthiness. You will design methods to systematically evaluate models against these standards, contributing to responsible AI deployment in policing and maximizing the efficiency of using open models. The research includes formalizing value taxonomies and alignment metrics, mapping operational requirements to formal criteria, and mathematically defining constructs such as 'trustworthiness' and 'fairness' for law enforcement workflows.

Engineering scalable methodologies for automated and human-in-the-loop evaluation and red-teaming pipelines is a key part of the project. You will explore red-teaming, synthetic data generation for vulnerability probing, and customize downstream alignment techniques (e.g., DPO, RLHF, constitutional AI) to enforce strict adherence to organizational values. The project is embedded in the Model-Driven Decisions Lab, a collaborative initiative between the Netherlands Police and TU Delft, where you will join an interdisciplinary community of PhD students. You will spend 20% of your time at the Netherlands Police’s strategy and innovation division and collaborate with the Delft Digital Ethics Centre at the Faculty of Technology, Policy, and Management (TPM). Your home base will be the Web Information Systems research group at the Computer Science faculty (EEMCS).

Applicants must hold an MSc in computer science, data science, or a relevant subject such as ethics of AI, with practical machine learning/artificial intelligence coursework and project/thesis experience. A strong interest in AI alignment, human-AI interaction, and explainable AI is required. Candidates must have excellent communication skills for interdisciplinary collaboration and a good command of written and spoken English. Due to the project context, a good command of the Dutch language is a strong requirement. Security screening by the Dutch government is mandatory before starting.

Employment conditions include a 5-year PhD contract (initial 1.5-year contract with progress assessment, followed by 3.5 years), salary from €3059 per month in the first year to €3881 in the fourth year, plus 8% holiday allowance and 8.3% end-of-year bonus. In the 5th year, a temporary monthly allowance is provided. Additional benefits include a customizable compensation package, health insurance discounts, monthly work costs contribution, and flexible work schedules. TU Delft offers support for relocation and integration into the Netherlands, including a Dual Career Programme for accompanying partners.

To apply, submit your application online via the provided link by 10 August 2026, addressing it to Dr. Ujwal Gadiraju. Applications sent by email or post will not be processed. English proficiency is required for doctoral education and scientific writing. For more information, contact Dr. Ujwal Gadiraju at the provided email address.