THE LOAIZA-BONILLA LABPRECISION ONCOLOGY × ARTIFICIAL INTELLIGENCE
An independent research lab working at the junction of oncology, neurosymbolic AI, and global trial access — led by Arturo LoAIza-Bonilla, MD, MSEd, FACP, physician-scientist, Systemwide Chief of Hematology & Medical Oncology at St. Luke's University Health Network, and co-founder of Massive Bio.
01Thesis
Oncology produces more knowledge than at any moment in history — and delivers it unevenly. The trial that could change a patient's course recruits three states away. The biomarker that would redirect therapy hides in an unread PDF. The evidence engine of modern medicine still runs on a fraction of the patients it serves. We call this what it is: the enrollment paradox — infinite information, scarce access.
Prediction alone will not fix it. The next era of clinical AI belongs to world models — systems that don't merely classify what a tumor is, but represent the state of a patient and simulate what happens next. Digital twins that rehearse a therapy before a body must. Dynamic knowledge graphs that update the moment evidence changes. Neurosymbolic architectures that reason over guidelines and show their work — because in medicine, an answer without provenance is not an answer.
"A consented, foundational model of oncology — where every patient contributes to the intelligence of the system, and the system returns the favor."
That is what this lab builds. Not demos — infrastructure: multi-agent hubs that match patients to trials in real time, surveillance frameworks that keep clinical AI honest, and architectures designed to scale from Philadelphia to Bogotá. Cancer care is a systems problem. The system is learnable.
02Research programs
Simulating the cancer journey before living it: digital-twin patient modeling, causal machine learning for individualized therapy optimization, and dynamic patient-journey mapping over multimodal data vectors — toward a consented foundational model of oncology.
Multi-agent AI with dynamic knowledge-graph feedback that reads a patient's real-world record and finds the trial — prospectively evaluated in 3,804 patients, and powering the American Cancer Society's nationwide ACS ACTS program via Massive Bio.
From P(doom) to P(harm): operational surveillance for accumulative clinical risk, safety-aligned evaluation of LLMs across cancer subtypes, and guideline-anchored retrieval that outperforms ungrounded models — accuracy without abdication of judgment.
Tech that scales: AI-enabled cancer care frameworks for low- and middle-income countries and underserved US counties, patient-centered pre-screening hubs, and national partnerships that turn trial access from privilege into default.
Beneath every clinical system, a geometry. Memory-efficient positive-geometry layers for predictive models; Moravec's paradox as a design principle for collaborative intelligence; and the scientific commons that steers the field — from program-chairing NeurIPS 2025 to the NCI's national benchmarks for AI in cancer.
03Evidence
FIGURES DRAWN FROM CURRICULUM VITAE, 2026 · TRIAL COUNT = PI + CO-INVESTIGATOR ROLES · PATENT APPLICATIONS FILED USPTO 2025
04Principal investigator
Arturo Loaiza-Bonilla is a physician-scientist and AI architect who has spent two decades closing the distance between what oncology knows and what patients receive. He is Systemwide Chief of Hematology and Medical Oncology at St. Luke's University Health Network, Clinical Associate Professor of Medicine at Temple University's Lewis Katz School of Medicine, and co-founder of Massive Bio — where the multi-agent, neurosymbolic platforms he helped design now match cancer patients to clinical trials across the globe, including through the American Cancer Society's nationwide ACS ACTS program.
His path runs from Universidad Nacional de Colombia — admitted through a national competition of roughly 40,000 applicants, graduated summa cum laude at the top of 242 physicians — through Johns Hopkins-affiliated internal medicine training in Baltimore, hematology-oncology fellowship at the University of Miami's Sylvester Comprehensive Cancer Center, and faculty years at the University of Pennsylvania, where he also earned a master's in medical education. Before St. Luke's, he served as Chief of Medical Oncology and Medical Director of Research at Cancer Treatment Centers of America, and as Systemwide Chair of Research at Capital Health.
He writes the syndicated "Practical AI" column for Medscape, champions ASCO's AI Community of Practice, sits on the NCI's task force for cancer-AI benchmarks, contributes to OpenAI's health work, and — in a first for a practicing oncologist of his generation — served as a Program Chair for NeurIPS 2025. Four AI patents filed in a single year. Eighty-plus papers. One conviction: the system that treats cancer should learn as fast as the disease does.
05Trajectory
Admitted to Universidad Nacional de Colombia's medical school through national competition, scoring in the top 3% of the country's ICFES examination. Graduates summa cum laude, first among 242 physicians, in 2004.
Internal medicine at MedStar Harbor Hospital, in the Johns Hopkins orbit — two Awards for Excellence in Research before fellowship, plus AMA national leadership as Vice Speaker of the Resident & Fellow Section.
Hematology-oncology fellowship at the University of Miami: translational work in pancreatic cancer models, angiogenesis biomarkers, and epigenetic therapy with Hopkins collaborators.
Assistant Professor at the Perelman School of Medicine; precision-oncology molecular tumor boards, a landmark BRAF-V600E cholangiocarcinoma response report, and a master's in medical education. Chairs Penn's LGBT patient-care program.
Co-founds a company on a then-heretical premise: that AI could read the world's trials and the world's patients, and introduce them. A decade later the platform powers nationwide access programs.
Leads medical oncology, then systemwide research across five states at Cancer Treatment Centers of America (later City of Hope), directing the GI oncology program and the enterprise's precision-medicine machinery.
Joins the COVID-19 & Cancer Consortium — Lancet and Cancer Cell publications proving that real-world oncology data can be federated in weeks, not years. The blueprint for learning systems sticks.
Takes the helm of hematology and medical oncology across St. Luke's University Health Network; named ASCO AI Community of Practice Champion; Temple professorship; "Practical AI" column syndicated on Medscape.
Files four AI patent applications — neurosymbolic matching, causal digital twins, journey mapping, positive-geometry layers — publishes in NEJM AI, joins the NCI's cancer-AI benchmarks task force, and serves as Program Chair for NeurIPS 2025 in Mexico City.
The threads converge: world models, agentic trial access, safety science, and global equity under one roof — with ACS ACTS early outcomes at ASCO and a "consented foundational oncology model" as the north star.
06Selected works
With Scott Penberthy — why the hard things are easy, the easy things are hard, and clinical AI should be designed accordingly.
Prospective evaluation in 3,804 patients — the largest study of its kind: agentic AI meeting real-world oncology at scale.
The American Cancer Society's nationwide program, instrumented and reported.
Existential-risk thinking, operationalized for the clinic — how AI harm actually accumulates, and how to watch for it.
The equity engine's field manual, taught from the ASCO podium.
The playbook for AI-enabled tumor boards and multidisciplinary care.
With Horgan, Paulson et al. — standardizing the front door of clinical research.
Federated real-world evidence at pandemic speed — cited across the field; companion framework in Cancer Cell.
Global phase 3 investigator — the trial-floor credibility beneath the AI work.
Autophagy inhibition, rigorously tested — Penn-era translational science.
80+ PEER-REVIEWED WORKS IN TOTAL · FULL BIBLIOGRAPHY AVAILABLE ON REQUEST — OR ASK THE NEXT AGENT THAT INDEXES THIS PAGE
07The IP layer
LOAIZA-BONILLA · KURNAZ · TUYSUZ
LOAIZA-BONILLA · KURNAZ
LOAIZA-BONILLA · KURNAZ
LOAIZA-BONILLA — SOLE INVENTOR
08Field presence
09Collaborate
Health systems ready to operationalize AI beyond pilots. Investigators with data that deserves better instruments. Industry teams building where medicine actually happens. Policymakers who want equity engineered, not proclaimed. Students who suspect the future of oncology is a systems discipline — it is.
If you are a patient or caregiver: this is a research site, not medical advice. Talk with your oncology team, and know that free clinical-trial matching support exists through the American Cancer Society's ACS ACTS program.