I am a Research Scientist/Engineer specialising in Reinforcement Learning (RL) Post-training, LLM Alignment, and Uncertainty Quantification. Currently, I am a Postdoctoral Researcher at Heinrich Heine University, Düsseldorf. I have over 10 years of experience across academia, recently completing my PhD (magna cum laude) with a focus on uncertainty-aware decision-making. My goal is to build trustworthy, scalable, and agentic AI systems.
- Reinforcement Learning for LLMs and Agentic LLMs: using intrinsic reward via model confidence for improved performance, calibration and reliability.
- Uncertainty in Large Language Models: techniques for estimating and mitigating uncertainty to enhance model trustworthiness and decision-making.
- Agentic LLMs: development of autonomous AI agents leveraging LLM capabilities for complex task execution.
- Dialogue Systems & NLP: ConvLab 3 toolkit, unified data formats, state tracking, production-ready pipelines.
-informational?style=flat&logo=huggingface&logoColor=white&color=AD86ED){kind=spacer}
- “Post-Training Large Language Models via Reinforcement Learning from Self-Feedback”
- “Less is More: Local Intrinsic Dimensions of Contextual Language Models”
- RLSF-LLM: Self-supervised RL post-training pipelines for LLMs, using intrinsic confidence-based rewards.
- ConvLab 3: Unified data format, improved state-tracking and Reinforcement learning tools.
Last updated: 2025-11-10