About

I am a Ph.D. candidate in Computer Science at Trinity College Dublin and a Marie Skłodowska-Curie Fellow, currently a visiting researcher at Imperial College London hosted by Dr. Yingzhen Li.

My research interests are CUDA and GPU kernels, uncertainty estimation, stochastic processes and probabilistic modelling for large language models, and reinforcement learning. I care about making large models both reliable and fast: knowing what they don't know, and running efficiently on real hardware. Recent work covers probabilistic low-rank adaptation of LLMs, scalable Gaussian processes and weight sharing in Bayesian neural networks. At Imperial I work on stochastic processes for LLMs.

Before Trinity, I studied Artificial Intelligence at Eötvös Loránd University (Budapest) and Computer Science at Northeast Forestry University, and worked on model acceleration at Ericsson and on autonomous-driving perception at Zeekr.

• CUDA & GPU kernels
• Uncertainty estimation
• Stochastic processes & probabilistic modelling for LLMs
• Reinforcement learning

Publications

• NeurIPS 2026

  BrainTRACE: Tracing Longitudinal, Multimodal, and Volumetric Evidence in Brain MRI Clinical Reasoning

  Qizhen Lan, Mengchen Fan, Hang Zhang, Jingwei Duan, Moule Lin, Jialin Chen, Baocheng Geng, Xiaoqian Jiang

  Conference on Neural Information Processing Systems

• ICML 2026

  Bayesian-LoRA: Probabilistic Low-Rank Adaptation of Large Language Models

  Moule Lin, Shuhao Guan, Andrea Patanè, David Gregg, Goetz Botterweck

  International Conference on Machine Learning

• AAAI 2026

  Flow-Induced Diagonal Gaussian Processes

  Moule Lin, Andrea Patanè, Weipeng Jing, Shuhao Guan, Goetz Botterweck

  AAAI Conference on Artificial Intelligence

• ICLR 2026

  Teaching VLMs to Admit Uncertainty in OCR from Lossy Visual Inputs

  Shuhao Guan, Moule Lin, Cheng Xu, Jiaqi Zhao, Derek Greene

  International Conference on Learning Representations

• ACL 2025

  PreP-OCR: A Complete Pipeline for Document Image Restoration and Enhanced OCR Accuracy

  Shuhao Guan, Moule Lin, Cheng Xu, Xinyi Liu, Jiaqi Zhao, Jiexin Fan, Qi Xu, Derek Greene

  Annual Meeting of the Association for Computational Linguistics (Main)

• AISTATS 2025

  Stochastic Weight Sharing for Bayesian Neural Networks

  Moule Lin, Shuhao Guan, Weipeng Jing, Goetz Botterweck, Andrea Patanè

  International Conference on Artificial Intelligence and Statistics

• EMNLP 2025

  Effective Synthetic Data and Test-Time Adaptation for OCR Correction

  Shuhao Guan, Cheng Xu, Moule Lin, Derek Greene

  Conference on Empirical Methods in Natural Language Processing (Main)

• IEEE TGRS

  Gaussian-Based Swap Operator for Context-Aware Extraction of Building Boundary Vectors

  Moule Lin, Weipeng Jing, Wenjie Zou, Zhiyang Qiu, Chao Li

  IEEE Transactions on Geoscience and Remote Sensing

Experience

• Visiting Researcher · Imperial College London 2026.10 – present Department of Computing, London, UK · hosted by Dr. Yingzhen Li. Research on stochastic processes for large language models.
• Marie Skłodowska-Curie Fellow · Trinity College Dublin 2024.04 – present Dublin, Ireland. Uncertainty estimation for large language models.
• Researcher · Ubotica Technologies MSCA secondment Ireland. Deploying uncertainty-aware models on satellites for reliable onboard prediction and decision-making, applying Bayesian learning and uncertainty estimation research to Ubotica's models.
• Researcher, Autonomous Driving · Zeekr (Geely Group) 2023.07 – 2023.11 Shanghai, China. Researched state-of-the-art BEV models for automatic parking with fisheye cameras and LiDAR; improved large-scale training efficiency by ~8% via prefetched data loading on 2M+ images and videos; built a low-light and dirty-camera detection model from data collection to deployment (ONNX, TensorRT) on the Zeekr vehicle platform.
• Researcher, Model Compression · AI Lab, Eötvös Loránd University 2022.07 – 2022.10 Budapest, Hungary. Researched pruning, distillation and sparse architectures; proposed a novel “Swap” structure with ~10% inference speed-up, co-authoring Optimized Vectorizing of Building Structures with Swap.
• Researcher, Model Fusion · Ericsson 2021.07 – 2021.10 Budapest, Hungary. Implemented new algorithms in CUDA/C++ (e.g. a Shift operator to accelerate inference); adapted Transformers into Swin-style local-attention variants for multi-modal data; analysed communication-network data to predict user behaviour and preferences.

Education

• Ph.D. in Computer Science · Trinity College Dublin 2024 – 2027 School of Computer Science and Statistics, Dublin, Ireland
• M.Sc. in Artificial Intelligence · Eötvös Loránd University 2021 – 2023 Faculty of Informatics, Budapest, Hungary · Stipendium Hungaricum Scholar · Excellent Thesis
• M.Sc. in Computer Science · Northeast Forestry University 2020 – 2024 College of Computer and Control Engineering, Harbin, China · National Graduate Scholarship (2022)
• B.Eng. in Software Engineering · Tianjin Agricultural University 2015 – 2019 Tianjin, China · Honorary Graduate

Honors & Awards

• Marie Skłodowska-Curie Fellowship, European Commission
• Stipendium Hungaricum Scholarship (full scholarship), Hungarian Government
• National Scholarship for Graduate Students, Ministry of Education of China, 2022
• First Class Postgraduate Scholarship (three times), Northeast Forestry University
• National First Prize, Lanqiao (Blue Bridge Cup) Programming Contest — ranked 27 / 51,282
• First Prize, Lanqiao (Blue Bridge Cup) Programming Contest, Tianjin Region
• Tianjin Municipal Government Scholarship

Notes & Blog