Home

Meisam Ashraf

Industrial AI Research Scientist

Time-Series Learning • Generative Modeling • Trustworthy ML for Medical Devices

I build machine learning systems that transform noisy physiological signals into reliable clinical decisions. My work bridges deep learning, probabilistic modeling, and real-world deployment in regulated environments.

---

Focus Areas

Learning from Clinical Time-Series

Robust modeling of ECG, PPG, invasive blood pressure, and multimodal CPR signals under noise, artefacts, and distribution shift.

Generative & Probabilistic Modeling

Diffusion models, uncertainty quantification, and symmetry-aware representations for scientific data.

Reliable ML in High-Stakes Systems

Validation, interpretability, cross-modal fusion, and deployment-aware research.

---

Selected Work

Defibrillation Outcome Prediction

Transformer-based multimodal models predicting shock success from ECG and invasive signals in real-time clinical workflows.

Annotation & Signal Intelligence Platform

Designed scalable annotation infrastructure to support PMCF studies and structured signal validation.

Generative Scientific Modeling

Research on diffusion-based and symmetry-aware generative architectures for structured scientific data representation.

---

Download CV | LinkedIn | GitHub | Contact