Alisa Sheinkman
ML researcher.
Machine Learning Engineer and Researcher with a PhD in Statistics and strong proficiency in Python, PyTorch, TensorFlow and JAX. Experienced in designing efficient learning schemes and implementing complex neural networks. Passionate about building reliable probabilstic ML systems and bridging the gap between theory and production-ready code.
During my PhD, I was working under the supervision of Sara Wade and developed advances in Bayesian deep modelling. Namely, I studied efficient inference schemes with a focus on scalable variational inference algorithms such as stochastic and black box variational inference. My thesis addresses the challenge of architecture specification of Bayesian neural networks, Bayesian model choice and model combination in the realms of big data and overparametrized deep models.
Publications
Probabilistic inference in Bayesian neural networks, A. S., PhD dissertation, 2025.
SMRS: advocating a unified reporting standard for surrogate models in the artificial intelligence era, E. Semenova, A. S., T. J. Hitge, S.M. Hall, J. Cockayne, NeurIPS, 2025 (accepted, to appear).
Understanding the Trade-offs in Accuracy and Uncertainty Quantification: Architecture and Inference Choices in Bayesian Neural Networks, A. S., S. Wade, in Proceedings of the ECML PKDD, 2025. Corresponding code.
Variational Bayesian Bow tie Neural Networks with Shrinkage, A. S., S. Wade, arXiv:2411.11132, 2024. Corresponding code.
Deep learning techniques for noise annoyance detection: results from an intensive workshop at the alan turing institute, A. Mitchell, E. Brown, R. Deo, Y. Hou, J. Kirton-Wingate, J. Liang, A. S., C. Soelistyo, H. Sood, and A. Wongprommoon The Journal of the Acoustical Society of America, 153(3_supplement): A262-A262, 2023.