I am a final-year PhD candidate in Computer Science at Radboud University, where I train, build, and evaluate LLM systems with two focuses: LLM agents (deep research agents, agentic search, tool use) and uncertainty quantification. My goal is to build reliable AI systems that solve complex information-seeking and reasoning tasks.

Alongside my research, I work with industry on applied LLM problems. Most recently, at Thomson Reuters Labs, I built a multi-agent search system for legal research and a controller for deep research agents based on trajectory-level signals, improving both the search trajectory and final task performance (EMNLP 2026). Before that, at KPN, I brought uncertainty estimation to a customer-facing RAG chatbot.

I work hands-on with LLM fine-tuning (full, LoRA), RL post-training (RLVR), retrieval and reranking, synthetic data generation, and benchmark construction. My work has been published at ACL, EMNLP, COLING, SIGIR, SIGIR-AP, ICTIR, and ACM Computing Surveys.