🧪Yet Another ICU Benchmark: a holistic framework for the standardization of clinical prediction model experiments. Provide custom datasets, cohorts, prediction tasks, endpoints, preprocessing, and models. Paper: https://arxiv.org/abs/2306.05109

Repository for the Paper: „On the Importance of Step-wise Embeddings for Heterogeneous Clinical Time-Series“

SA-AKI Mortality Prediction — Survival analysis & binary classification for Sepsis-Associated Acute Kidney Injury using MIMIC-IV. CatBoost, LightGBM, XGBoost, Logistic Regression. AUROC ~0.80+

Early sepsis risk prediction pipeline using machine learning on ICU clinical data.

Hospital readmission prediction ML pipeline in R — 6 algorithms (LASSO, Random Forest, SVM, KNN, Naive Bayes, Decision Tree) on 69,984 diabetic patients using Tidymodels + SMOTE

Predicts 30-day hospital readmission risk in diabetic patients using ICD-9 code grouping, random oversampling, and HistGradientBoosting on 101,766 clinical encounters. Best model: F1 = 0.2818, AUC = 0.6704.

Stroke risk prediction using Random Forest in R — ROSE class balancing, 14.03% OOB error rate, confusion matrix evaluation & deployed as .rds model for clinical use

Reproducible CPU-only clinical risk prediction MLOps platform with calibration, conformal abstention, fairness, drift monitoring, and CI/CD.

ETL pipeline converting Synthetic K-MIMIC (SYN-ICU) to MEDS format, with a transportable in-hospital mortality benchmark. Submitted to SD4H Workshop @ ICML 2026.

Cardiac risk prediction Streamlit app — XGBoost/RF on Cleveland dataset biomarkers, calibrated probability output, SHAP attribution, ROC curves, AUC-optimised threshold.

Ethical approach to mental health screening using privacy-preserving machine learning for clinical decision support

Production ML pipeline for clinical decision support - risk scoring, treatment optimization, outcome prediction

🔬 Binary classification of benign and malignant breast tumours

Machine learning pipeline for multi-class treatment prediction in lung adenocarcinoma (LUAD) using patient-level molecular profiles, featuring ensemble-based model aggregation, benchmarking across diverse classifier architectures, and systematic performance evaluation.

End-to-end ML demo for clinical risk scoring on synthetic tabular patient data (data → features → models → metrics → FastAPI /predict).

End-to-end ML pipeline for predicting 30-day hospital readmission in diabetic patients — 8 models, SHAP explainability, clinical interpretation | UCI 130-US Hospitals dataset

Foundation model for health trajectory prediction — architecture shootout across Transformer, Mamba, and Continuous-Time

ICU in-hospital mortality prediction using MIMIC-III — classical ML, survival analysis, and deep learning on the first 24 hours of an ICU stay.

Multi-age ASD screening ML pipeline — AQ-10 · SHAP · Gradio · Uncertainty Quantification

Evaluates data efficiency in lung cancer risk prediction using a super-stacking ensemble. Trains models on progressively reduced fractions of the PLCO dataset while keeping a fixed test set, analyzing performance stability, degradation, and robustness under limited data.