Soumyadeep Roy

I am a postdoctoral scholar working with Prof. Tina Hernandez-Boussard at the Division of Computational Medicine, Department of Medicine, Stanford University. My current work is in perioperative pain management, alongside a clinical team of surgeons, trauma physicians, and psychologists, on two fronts: developing auditable reasoning LLMs that ground clinical recommendations in established guidelines, and using LLMs to construct executable guideline pathways that reveal when real-world patient trajectories deviate from them — and whether those deviations help or harm.

My work sits at the intersection of biomedical AI, machine learning, natural language processing, and real-world evidence. As medicine continues to generate increasingly complex clinical, genomic, and textual data, this kind of research is central to the future of AI in medicine.

Worked with clinical data (structured EHR and unstructured notes) of Parkinson’s disease (L3S Research Center, Hannover Medical School, Germany), oncology (breast, lung and prostate cancer) at GE Healthcare India and postoperative pain management at Stanford Medicine.

Translational Works

[Boston GrandHack 2026] Co-developed AuriCare, a holistic pain-management decision-support concept, presented at MIT Hacking Medicine’s Boston GrandHack 2026. Blog Event

[Patent, filed 2025] Lead co-inventor on a deep-learning representation-learning system for medical-imaging-equipment sensor logs, enabling anomaly detection and predictive maintenance (Wipro GE Healthcare).

[Translation Funded Project] My PhD work on interpretable clinical trial search has been continued as a funded translational research project at AI4ICPS programme at IIT Kharagpur, led by my co-author Official Website Poster