NexusMed.ai — ADR Outcome Prediction with Advanced Information Retrieval

Our Research

NexusMed.ai works to understand, predict, and prevent Adverse Drug Reactions (ADRs), bringing together large-scale pharmacovigilance data, advanced information retrieval, and clinical expertise. Our central focus is structured and unstructured ADR outcome predictions — forecasting the severity and clinical trajectory of adverse events at System Organ Class (SOC) level — where we confront the challenges of unstructured case narratives and the limitations of today's benchmarks.

We leverage retrieval as a first-class component of prediction. Rather than relying on model parameters alone, we ground every prediction in evidence drawn from real-world safety reports. Our mission is to strengthen patient safety and support informed prescribing by surfacing the hidden patterns within drug safety data.

Current Projects

Pharmacovigilance Database Mining Mining large-scale safety databases (FAERS, MedDRA, UMLS, PubChem) to detect and predict ADRs across both structured records and unstructured narratives.

Multi-Modal Retrieval Infrastructure Developing scalable BM25, dense vector, and GraphRAG databases that integrate lexical, semantic, and relational evidence for explainable pharmacovigilance analytics

Hybrid Retrieval Architecture Designing retrieval-augmented pipelines that combine lexical, semantic, and graph-based search to surface the most relevant ADR evidence for downstream AI models.

MCP-Powered Deep Research Agent Building an agentic system on fine-tuned LLMs that re-ranks retrieved evidence and refines queries during inference and content generation.

Natural Language Knowledge Exploration Supporting both natural-language and expert-mode queries to interactively explore over two decades of FAERS data at scale.

Real-Time Clinical Intelligence Developed a patient-centric, production-ready interface with a portable deployment framework, enabling real-time ADR reporting and interactive visualizations to enhance clinical decision-making.

Our Team

Dr. David Guo Principal Developer, Dual PhDs in IT and Biochemistry AI/ML in Healthcare and FinTech

Dr. Nishant Vishwamitra Assistant Professor, Information Systems and Cyber Security Image privacy, crowdsourcing

Dr. Kim-Kwang Raymond Choo Professor, Information Systems and Cyber Security; Cloud Technology Endowed Professorship III Blockchain, AI in cybersecurity

Collaborations

We are looking for partners with academic groups, clinicians, and regulatory-science teams on drug safety and retrieval-augmented AI. We welcome new collaborations across pharmacovigilance data sharing, clinical validation, and methods research. If you would like to be a part of these efforts, please get in touch.

Selected Publications

David Guo, Nishant Vishwamitra, and Kim-Kwang Raymond Choo. Attention-Weighted Retrieval-Augmented Data Engineering for ADR Severity Classification. EMNLP Industry Track, 2026. Github Repo
David Guo and Kim-Kwang Raymond Choo. Applications of Federated Large Language Models for Adverse Drug Reactions Prediction: A Scoping Review. Journal of Medical Internet Research, 2025.

Demo

Explore Retrieval-Augmented Knowledge Graph for ADR Prediction Framework (Access Upon Request)

GraphRAG Project

Contact Us

Email: contact@nexusmed.ai

Affiliated Organization: College of AI, Cyber and Computing @ UTSA

Donations

NexusMed.ai is a non-profit academic research organization sustained by donations, which primarily fund the computing and open-source development behind our projects. We welcome support in many forms, including cash, computing hardware (e.g., GPUs), and cloud credits. Ongoing contributions from a broad base of supporters are essential to our progress and to maintaining our public charity status. To learn more about sponsorship opportunities and benefits, please contact Dr. David Guo.

Predict the Harm.Prevent the Outcome.