I am Chief Applied Scientist for Health AI Innovation at Oracle Health, where I lead the science behind AI products for population health and the life sciences, from clinical trial matching to evidence at the point of care. My teams have taken more than twenty AI systems from idea to general release, spanning predictive, causal, generative and agentic models. I care as much about proving these tools work as about building them, so much of my time goes into the evaluation and monitoring frameworks that tell us whether an AI system is safe, fair and delivering the outcomes it was designed for.
I am an academic data scientist and health services researcher by background, and I have dedicated my career to designing, deploying and evaluating computational techniques that help us make better decisions about health and healthcare. My academic career was driven by three core notions: that patient-centered data can provide unique insights into patient care, outcomes, and experience; that novel data can be burdensome to collect; that systems can be built to reduce the burden of collection by providing strong utility at the point of care.
For example, my labs have shown that patient-reported outcome measures (PROMs) can be used as a clinical intervention to improve quality of life and identification of need, and even patient survival. We have consistently shown that it is possible to dramatically reduce the burden of PROMs without sacrificing accuracy by using computerized adaptive testing (e.g., Harrison and Gibbons, 2022). We are now developing agent-derived assessments (ADAs), which use AI agents and machine learning models to add to the information patients share through PROMs. As data science and AI lead for the Lancet Commission on Cancer and Health Systems, I led work using LLMs to assess and report the burden of cancer, creating actionable insights for policymakers around the world. Through my post-doctoral Fellow and long-term collaboarator Dr. Andre Pfob, this work has recently received $20m from the European Union.
Our award-winning collaborative research on AI-enhanced techniques to identify pathologic complete response (pCR) after neoadjuvant chemotherapy for localized breast cancer suggests that many patients may be able to safely avoid confirmatory surgery (e.g., J Clin Oncol, Eur J Cancer).
We strive to build effective, safe and unbiased AI tools that inform clinical decisions to improve outcomes, reduce costs, and ensure that appropriate care is provided at the end of life. With Project Ronin, we showed how these tools can be implemented fairly and safely in real-world care. I chaired the Coalition for Health AI (CHAI) working group on model cards, which helps set national standards for how health AI models are documented and disclosed.
At MD Anderson, I founded the Section of Patient-Centered Analytics. I also designed and ran the Institute for Data Science in Oncology Fellowship, which gives postdoctoral researchers and junior medical staff the skills and critical insight to become leaders in oncology data science. Before that, I moved from Cambridge to Brigham and Women’s Hospital and Harvard Medical School as founding Co-Director of the Patient-Reported Outcomes, Value & Experience (PROVE) Center, which now serves as the Central hub for patient-reported data and analytics across the Mass General Brigham ecosystem.
I maintain links with the University of Cambridge as Director of Health Assessment and Innovation at the Psychometrics Centre and Health Director of the Concerto Platform. I was previously a National Institute for Health Research (NIHR) Fellow at The Healthcare Improvement Studies (THIS) Institute and a College Research Associate at Corpus Christi College, Cambridge.
I have been invited to lecture on data science and patient-reported outcomes at institutions including NASA, Karolinska Institutet and Harvard Business School. From 2014 until 2022, I taught in-person courses on modern psychometrics, big data and machine learning at the University of Cambridge.