Selected Projects
A selection of clinically grounded AI systems and data science projects, focused on medical imaging, risk stratification, and decision support.
TrustSyn | Biomedical AI Research
A biomedical machine learning research prototype exploring anticancer drug-combination activity prediction using NCI-ALMANAC experimental data, with split-specific evaluation and model interpretability.
- Drug-combination activity prediction
- CatBoost, XGBoost, D-MPNN and stacked models
- Held-out evaluation and SHAP-based interpretability
- Planned API, containerization and cloud engineering
Anesthesia Drift Detector
A monitoring pipeline to detect distributional drift in anesthesia-related physiological signals, enabling early warnings when model assumptions degrade in real-world clinical settings.
- Time-series monitoring & drift detection
- Clinical safety & model reliability
- Python, statistical monitoring, visualization
HCC Extractor (Hierarchical Condition Category)
A hybrid NLP and rule-based system that derives ICD-10 diagnosis codes from unstructured clinical text and maps them to Hierarchical Condition Categories (HCCs) for risk adjustment and analytics.
- Pipeline: Clinical text → ICD-10 → HCC mapping
- NLP + rule-based extraction
- Healthcare coding & structured data generation
- Python, text processing, validation logic
PinkScan AI
A non-invasive breast health screening concept combining imaging signals with AI-driven decision support to enable scalable early screening workflows in resource-constrained settings.
- Clinical screening & triage support
- Imaging-based inference
- Model prototyping & evaluation