Parastoo is a visiting PhD researcher from the Machine Learning Group at TU Berlin / BIFOLD, where she is completing an industrial doctorate in collaboration with BASF on machine learning and explainable AI for inorganic material and catalyst design. Her research addresses the practical challenges of learning from experimental datasets with severe class imbalance and limited labeled samples, using interpretability methods — including layer-wise relevance propagation (LRP) and SHAP-based attribution — to translate model outputs into actionable guidance for experimentalists designing the next round of synthesis. At OptiML, she applies Bayesian optimization with Gaussian process surrogates to bioprocess control, developing sequential experiment design strategies that extract maximum information from costly fermentation runs. Her work has appeared in ACS (Editors’ Choice) and at NeurIPS workshops.