Journal Article · December 2025
Safe-By-Design Clinical LLMs: Formal Risk Models and Silent-Trial Protocols for Workflow Integration
Charan Tej Gayapu · Macaw International Journal of Advanced Research in Computer Science and Engineering · Vol. 11, No. 2, pp. 8–15
Addresses the integration of large language models into clinical workflows with an emphasis
on safety: formal risk modeling and silent-trial protocols that allow AI systems to be
evaluated in live clinical environments without influencing care, before workflow deployment.
Large Language ModelsClinical AI Safety
Risk ModelsSilent TrialsWorkflow Integration
Sample Entry · 2025
Transformer-Based Automated MedDRA Coding of Adverse Event Verbatim Terms
Sample publication — replace with your own paper details (authors, journal, DOI, citations)
Fine-tuned clinical language models for preferred-term assignment with a human-in-the-loop workflow for regulatory compliance.
NLPMedDRAPharmacovigilance
Sample Entry · 2024
Metadata-Driven Automation of SDTM Mapping: A Scalable Framework for Submission Readiness
Sample publication — replace with your own paper details
A reusable metadata repository and rules engine that generates SDTM datasets and define.xml, reducing mapping cycle time from weeks to days.
CDISCSDTMAutomation
Sample Entry · 2024
Linking EHR and Trial Data for Real-World Evidence Generation
Sample publication — replace with your own paper details
A privacy-preserving record-linkage methodology connecting EHR-derived endpoints to trial outcomes.
RWEEHRData Linkage
Sample Entry · 2023
Machine Learning–Enhanced Key Risk Indicators for Risk-Based Monitoring
Sample publication — replace with your own paper details
Gradient-boosted KRI models that predict site-level data quality issues earlier than threshold-based approaches.
RBMMachine LearningSite Quality
Sample Entry · 2022
Data Management Challenges in Decentralized Clinical Trials: Wearables, eCOA, and Beyond
Sample publication — replace with your own paper details
A framework for ingesting, standardizing, and quality-controlling high-frequency sensor data streams within CDISC-conformant pipelines.
DCTWearableseCOA