Research

Publications

Peer-reviewed research from Virtual Hipster AI and collaborators, spanning AI, bioinformatics, and drug discovery.

Peer-Reviewed Publication · Scientific Reports (Nature Portfolio)

RAG-based architectures for drug side effect retrieval using compact LLMs

Shad Nygren (lead author), Omer Erdogan, Pinar Avci, Andre Daniels, Reza Rassool, Afshin Beheshti & Diego Galeano. Scientific Reports 16, 12754 (2026). Published 9 March 2026.

The authors are affiliated with institutions including the Broad Institute of MIT and Harvard, Harvard Medical School and Massachusetts General Hospital (Wellman Center for Photomedicine), the University of Pittsburgh, Koç University (KUIS AI Center), the Universidad Nacional de Asunción, and Kwaai.

We evaluate retrieval-augmented architectures that inject curated knowledge from the Side Effect Resource (SIDER) into compact LLM workflows for answering drug-side-effect questions. The graph-based approach (GraphRAG) achieved near-perfect accuracy—returning exact results with 100% precision, recall, and F1 on reverse queries—at substantially lower latency than a text-based baseline.

Authors

Shad Nygren · Lead author

Kwaai · Founder, Virtual Hipster

AI & cloud solutions architect and founder of Virtual Hipster; AWS Certified Solutions Architect and AWS Certified Security – Specialty. Lead author on this work in retrieval-augmented architectures for drug side-effect retrieval.

Focus: AI Consulting · Cloud Architecture · Retrieval-Augmented Generation · AWS

ORCID · LinkedIn · GitHub

Omer Erdogan · Co-lead author

Koç University & KUIS AI Center, Istanbul

Computer-engineering and AI researcher at Koç University and the KUIS AI Center, Istanbul. Co-lead author (equal contribution), focused on the large-language-model and retrieval-augmented-generation engineering behind this work.

Focus: Large Language Models · Retrieval-Augmented Generation · Machine Learning · Computer Engineering

Pinar Avci, MD

Wellman Center for Photomedicine · MGH, Harvard Medical School & Semmelweis University

Dermatology and photomedicine researcher at the Wellman Center for Photomedicine (Massachusetts General Hospital / Harvard Medical School) and Semmelweis University. A highly-cited author whose work spans photobiomodulation, low-level laser therapy, skin, and hair loss and regrowth.

Focus: Photobiomodulation · Low-Level Laser Therapy · Dermatology · Hair Loss

Andre Daniels

Kwaai

Pharmacist contributing clinical pharmacology and drug-safety domain expertise—directly relevant to this work on drug side-effect retrieval—as a contributor at Kwaai, the open-source AI lab.

Focus: Pharmacy · Drug Safety · Pharmacology

Reza Rassool

Kwaai · Affiliate Professor, University of Washington

Founder and chair of Kwaai and a RealNetworks Fellow; former CTO of RealNetworks. A prolific inventor with 27 U.S. patents (75 WIPO)—largely in digital video and streaming media—plus Technical Oscar and Emmy awards; co-founder of Widevine (acquired by Google) and Affiliate Professor of Computer Science and Physics at the University of Washington.

Focus: Computer Vision · Edge AI · Streaming Media · Embedded Systems

Afshin Beheshti, PhD

University of Pittsburgh · Broad Institute of MIT and Harvard

Professor of Surgery and Computational & Systems Biology at the University of Pittsburgh; Director of the Center for Space Biomedicine and Associate Director of the McGowan Institute for Regenerative Medicine; visiting scientist at the Broad Institute of MIT and Harvard. PhD in Physics (Florida State University); research spans space biology, systems biology, radiation biology, and bioinformatics.

Focus: Space Biology · Systems Biology · Radiation Biology · Bioinformatics

ORCID · Google Scholar · ResearchGate · LinkedIn

Diego Galeano, PhD

Universidad Nacional de Asunción, Paraguay

Researcher in the Department of Electronics and Mechatronics Engineering at the Universidad Nacional de Asunción, Paraguay. Specializes in AI for drug discovery and machine-learning prediction of drug side effects; developed the geometric self-expressive model (GSEM) and co-authored work in Cell Reports Methods.

Focus: AI for Drug Discovery · Drug Side-Effect Prediction · Machine Learning · Computational Biology

Google Scholar · LinkedIn

Selected Publications — Co-author Pinar Avci, MD

Pinar Avci is a highly-cited photomedicine researcher (Wellman Center for Photomedicine, MGH & Harvard Medical School). Selected first-author works:

Low-level laser (light) therapy (LLLT) in skin: stimulating, healing, restoring

Seminars in Cutaneous Medicine and Surgery (2013)

~1,486 citations (Google Scholar)

Read →

Low-level laser (light) therapy (LLLT) for treatment of hair loss

Lasers in Surgery and Medicine (2014)

~466 citations (Google Scholar)

Read →

Animal models of skin disease for drug discovery

Expert Opinion on Drug Discovery (2013)

~167 citations (Google Scholar)

Read →