Maryam Bayat Makou

Define and implement data-driven solutions across multiple AI and analytics projects, leveraging large-scale, heterogeneous datasets. Perform data cleaning, validation, and structuring to ensure reliability and reproducibility of downstream analysis and models. Conduct feasibility and performance analyses to assess data quality and operational constraints. Develop reusable Python-based analytics tools for evaluation, reporting, and decision support. Apply explainable AI techniques (SHAP and custom methods) to improve model transparency and stakeholder trust. Collaborate with domain experts to align analytical solutions with system and hardware requirements.

Designed and delivered end-to-end data analysis pipelines on large, heterogeneous datasets. Built and optimized supervised machine learning models (XGBoost, neural networks), achieving a two-fold performance improvement through tailored handling of imbalanced data. Diagnosed and improved model behavior using explainable AI techniques, strengthening robustness and interpretability. Translated analytical results into clear, actionable insights for technical and non-technical audiences.

Developed and optimized deep learning models in Keras/TensorFlow for rare-event classification, with a focus on interpretability via Taylor expansion. Improved model convergence using PCA-based decorrelation and robustness-focused feature preparation.