Oskar Kviman

Oskar Kviman is a doctoral student at KTH Royal Institute of Technology working in the Lagergren Lab within the Division of Computational Science and Technology. His research bridges machine learning, statistics, and computational biology with a focus on developing and applying advanced probabilistic methods.

His primary research interests include:

• Bayesian phylogenetics and probabilistic machine learning
• Variational inference, variational auto-encoders, and sequential Monte Carlo methods
• Generative AI techniques including flow matching, Schrödinger bridges, and diffusion models
• Computational cancer research focusing on differential expression testing and spatial transcriptomics

Kviman's publication record demonstrates significant contributions to variational inference methodology, particularly in phylogenetics and generative modeling. His work spans top machine learning conferences including ICML, NeurIPS, and AISTATS, showing consistent development of techniques that improve efficiency and accuracy in probabilistic modeling. Recent publications focus on multi-marginal flow matching, variational resampling, and mixture learning in black-box variational inference.

He has been recognized for his peer review contributions as a Top reviewer (10%) for AISTATS 2023. Kviman has supervised master's theses for Xindi Liu and Ricky Molén at KTH and serves as a lecturer for 'Statistical Methods in Applied Computer Science' since 2021, while previously working as a teaching assistant for 'Machine Learning, Advanced Course' and 'Deep Learning, Advanced Course'.