model-interpretability

Powerful Python tool for visualizing and interacting with pre-trained Masked Language Models (MLMs) like BERT. Features include self-attention visualization, masked token prediction, model fine-tuning, embedding analysis with PCA/t-SNE, and SHAP-based model interpretability.

Python project exploring interpretable and explainable AI methods for logic-based decision models.

End-to-end Machine Learning + MLOps project for predicting student math scores using multiple regression models, MLflow experiment tracking, SHAP interpretability, and Streamlit deployment.

Universal visualization library for N-dimensional machine learning decision boundaries. Works with PyTorch, Keras, Sklearn, and any custom model or dimensionality reducer (PCA, UMAP, t-SNE, and more) to render interactive 2D/3D plots.

Cost-sensitive loan default prediction using Python and machine learning, with threshold optimization, business cost simulation, model interpretation, and responsible AI considerations.

Machine learning regression model predicting 1985 automobile prices. Lasso model achieves 91.7% R² with superior generalization over XGBoost. Handles extreme multicollinearity (VIF 16,676→8.36), data leakage detection, and outlier treatment through PCA and domain-driven feature engineering.

Machine learning pipeline using Gradient Boosting to forecast regulatory compliance risk with explainable AI techniques (SHAP, PDP) on QuantGov data.

End-to-end anomaly detection system for financial transactions using machine learning and deep learning techniques.

Federated Learning Simulation on a Single GPU with Model Interpretability and Interactive Visualization

A lightweight Explainable AI CNN for PathMNIST medical imaging, achieving 91%+ accuracy with Integrated Gradients and SQLite-based attribution storage. Built in PyTorch, this scalable model delivers high performance, transparency, and real-world readiness, making it ideal for medical AI, edge deployment, and explainable deep learning research.

Explainable AI (XAI) based system for detecting financial fraud using machine learning, with model interpretability, analysis, and research-backed implementation.

Employee performance prediction using XGBoost multiclass classification (92.5% accuracy, 93.3% CV F1-score) with SHAP interpretability. Analyzes 1,200 employee records across 28 features, identifies top 3 performance drivers, and provides HR recommendations. Full pipeline: EDA, feature engineering, model comparison, and deployment-ready inferen

hybrid-talent-rec-benchmarks

Model interpretability for Explainable Artificial Intelligence

Customer churn prediction system using XGBoost, SHAP explainability, and Streamlit for real-time telecom retention analysis.

Leakage-safe comparison of Linear, Ridge, Lasso, and Elastic Net regression for used-car price prediction with cross-validation, coefficient stability, and multicollinearity analysis.

Research scripts and dashboards analyzing linguistic steering and Shapley-value attribution in language-model experiments.

A visual analytics tool and framework for exploring compositionality in sentence embeddings. Gain interactive insights into how embedding models, composition functions, and similarity metrics influence textual representations, focusing on error gap analysis for enhanced model interpretability.

Mechanistic interpretability workbench for circuit tracing, causal interventions, attribution comparison, and hypothesis validation.