Creating interpretable and transparent AI by bridging machine learning and neuroscience

Bahar Tolooshams is a leading expert in interdisciplinary AI. As the founder and director of the NeuBahar Lab, her research focuses on developing Neuro-Bayesian AI for human-interpretable abstractions and representation learning, bridging machine learning with neuroscience.

Tolooshams's work is at the forefront of creating AI systems that are both effective and transparent, with a particular emphasis on applications in speech enhancement, neural signal analysis, and medical imaging. Her publications, featured in top-tier conferences and journals like Transactions on Machine Learning Research and Neuron, demonstrate a deep expertise in dictionary learning, deep residual autoencoders, and deep learning for deconvolutional analysis. This research is instrumental in advancing AI's capability for robust, stable, and interpretable signal processing and data analysis.

Before joining the University of Alberta, she completed a PhD at Harvard University and held the Swartz Foundation Postdoctoral Fellowship at Caltech's AI for Science Lab. Her expertise is further solidified by her industry experience at Amazon AI and Microsoft. Beyond her research, Tolooshams is dedicated to community building, having actively mentored students through programs at Harvard and through peer-to-peer support networks.