The Cao Lab for Medical AGI/ASI builds medical artificial general and superintelligence to push the frontiers of medical knowledge and make world-class healthcare available for all.

Established in 2024, the lab’s mission is to elevate humanity's collective medical wisdom and create a healthier and more equitable world through the development and application of medical AI systems.

Clarity

Bryan Lee, Dr. Deepak Rajan-Jeyarajan, Chun Joo Goh, Keren Collins, Qiao Er Tee

Clinical guidelines remain underutilised in Australian healthcare settings. This project aims to develop an OpenEvidence-inspired, AI-enabled platform to make proprietary and open-source guidelines more accessible, interpretable, and actionable in everyday clinical practice.

Cybersecurity

Chun Joo Goh, James Liu

Cybersecurity risks associated with medical large language models are not yet well understood. This project aims to produce a clear, practical explainer for clinicians and hospital IT teams on the cybersecurity challenges of deploying LLMs in clinical settings, along with evidence-based mitigation strategies.

Economic Index

Izabella Mancewicz, Yufei Xu

The extent and nature of large language model (LLM) use across medical specialties remain poorly understood. This project aims to examine how physicians in different specialties utilise commercial LLMs through analysis of the Anthropic Economic Index, a dataset capturing millions of interactions with Anthropic’s Claude platform.

Sleepwell

Bryan Lee, Ryan Kua, Dr. Deepak Rajan-Jeyarajan, Dr. Nethum Devendra

Sleep quality among hospital inpatients is often poor. This project aims to evaluate whether the use of earplugs can meaningfully improve sleep quality on hospital wards.

Spectrum BPD

Dr. Deepak Rajan-Jeyarajan, Chun Joo Goh, Keren Collins, Qiao Er Tee

Patients with personality disorders in Victoria often receive structured support during clinic appointments but have limited continuity of care between visits. This project aims to develop an AI-enabled support system to help bridge this gap and provide safe, clinically informed assistance outside scheduled appointments.

State of Medical AI

Lucy Lu, Izabella Mancewicz, Yufei Xu

We currently lack a comprehensive picture of how Australian health services are adopting and integrating artificial intelligence. This project aims to survey health services nationwide to understand how AI is being implemented in practice and how it is incorporated into broader organisational strategy.

Swifts Creek

Keren Collins, James Liu

Bush nurses in Australia serve as the primary point of care for a large proportion of the rural and remote population, yet often operate with limited clinical and institutional support. This project aims to develop an AI-enabled assistant to enhance decision-making, access to information, and continuity of care in these settings.

TGA Commentary

Lucy Lu, Izabella Mancewicz, Yufei Xu

From a regulatory standpoint, Australia is at a crossroads in deciding whether to follow the United States in relaxing regulations for medical AI or the European Union in strengthening them. This project seeks to explore a third, more nuanced path. Assuming the TGA’s funding and capacity remain unchanged, how can regulatory burden be reduced while strengthening clinical safety and enabling innovators to develop AI responsibly?

Deepneuron

Keren Collins, Arsa & others from MDN

It remains uncertain whether current AI systems are sufficiently advanced to perform even relatively straightforward surgical procedures. This project aims to develop a simulated surgical environment for the appendix and to train a surgical AI model, using reinforcement learning, to perform an uncomplicated appendicectomy.

Bloomed

Bryan Lee, Dr. Deepak Rajan-Jeyarajan, Felix, Arsa

Medical student education remains inefficient and highly variable in its ability to identify and address individual learning gaps. This project aims to use AI to actively map conceptual weaknesses and support the development of structured, durable mastery.

NudgeAI Buntine

Dr. Deepak Rajan-Jeyarajan, Ryan Kua, James Liu

Unnecessary diagnostic testing remains common in emergency care, contributing to increased costs, patient burden, and system inefficiency. Building on initiatives such as the No Unnecessary Tests (NUTS) program led by Safer Care Victoria and Professor Paul Buntine, this project aims to develop an AI-enabled nudge-based system to support clinicians in reducing low-value testing while maintaining clinical safety.

MARC Projects

Chun Joo Goh/Keren Collins, Qiao Er Tee

MARC is a curated portfolio of projects arising from an annual call for proposals from physicians for medical AI collaborations. All projects involve AI/automation and include: Automated discharge summaries (Alfred, Anton Peleg/Aadith Ashok) Diagnostic reasoning for fever of unknown origin (Alfred, Anton Peleg/Aadith Ashok) Occupational history assistant with ISCO-88 encoding (Alfred, Ryan Hoy) Allied health search engine for dementia patients (Alfred, Marianne Coleman) IBD severity classification from imaging (Eastern, Abhinav Vasudevan) Cochlear implant candidacy assessment (Ear Science Institute Australia, David Sly) Outpatient triage for gastroenterology and neurology (Monash, Darcy Holt, Thanh Phan) Follow-up scheduling from scope reports (Peninsula, Kim Hay Be) Consultant-level feedback for emergency registrars (Peninsula, Gabriel Blechler) Dementia risk prediction from electronic medical records (Peninsula, Alicia Lu) Admission-based ICU prognostication (Peninsula, Nilesh Shah)

HARP Projects

Dr. Nethum Devendra, James Liu, Ryan Kua, Mark, Viran

HARP is a curated portfolio of projects arising from an annual call for proposals from physicians for medical hardware collaborations (currently launching). Projects include: Magnetic-guided gastrojejunostomy formation (Alfred, Larry Lai) 3D-printed models for wisdom tooth extraction training (Independent Dentist, Yulin Weng) (⁠wisdom-trainer)