The urge to make things has been with me for as long as I can remember. It has carried me through every business I have started, and through a career rooted in one plain goal: helping patients and providers collaborate more effectively. I came to it early — partly through clinical-care work in the United States, partly through time in a Finnish hospital, where I saw what becomes possible when the relationship between patient and provider is treated as the substrate of care rather than a byproduct of it. I came home with those observations, and I have spent thirty-nine years acting on them.
What I do, stated plainly: extract clinical wisdom from experts and encode it into engaging, compassionate programs for people living with chronic illness and the providers caring for them. The applications change with the era — expert systems in the 1980s, web platforms in the 2000s, AI partnerships now. The commitment underneath has not moved.
Making that technically real meant taking what I call the semantic path. In plain terms: hold each person as a coherent picture — their goals, their circumstances, their patterns — rather than as a scattered set of preferences and clicks. Technically it means building structured representations, ontologies, that hold the knowledge universe of clinical care and behavior in a form a machine can reason over. Most personalization in healthcare technology is preference settings and user-data records. I model each person as a context-aware semantic entity — interests, goals, traits, constraints — mapped to domain ontologies that develop with use. That commitment shaped a 2017 platform architecture which still describes, to my mind, what a system encoding clinical wisdom needs to be.
The chronology that follows is 39 years of that mission, played out across different companies, technologies, and clinical domains. The mission has not changed. The tools have.
My interest in the convergence of healthcare and artificial intelligence dates to the early 1980s — years before the term e-health existed and more than a decade before the internet would reshape medicine. In 1987 I co-founded Health Information Technologies, Inc. to pursue it, with Jonathan D. Lieff, M.D., a Harvard-trained neuropsychiatrist and past president of the American Association for Geriatric Psychiatry. The work I did there in expert systems and clinical decision support put me among a small cohort of practitioners who thought the future of healthcare would be built on structured knowledge, not just data.
By 1992, this conviction had produced the Medical Behavioral Management System (MBMS) — a multi-component clinical intelligence platform that included a treatment algorithm, a self-learning intelligence accumulator, and an automated counselor that delivered personalized patient guidance between visits. The MBMS architecture anticipated what the field now calls collaborative intelligence: the structured integration of human clinical expertise with AI-driven decision support. Raymond Kurzweil advised the company from 1993.
The question was never whether AI could replace clinicians. It was whether we could build systems intelligent enough to learn from them.
At the core of my work is a focus that remains rare in health technology: behavioral informatics — understanding and influencing human behavior to reach better health outcomes. Most digital health initiatives optimize workflows or digitize records. I start from a different question: how do people actually change?
Two projects from the 1990s established the methodology in practice. Talk From The Heart (1997) applied it to hypertension, sponsored by Merck and built with Dr. Donald E. Morisky of UCLA — whose Morisky Medication Adherence Scale remains one of the most widely used adherence instruments in the world. Diabetes Talk (1998) applied it to Type 2 diabetes patients transitioning from oral medications to insulin, sponsored by Eli Lilly and built with Dr. Richard Rubin, named ADA's Outstanding Educator in Diabetes for 1997. Both ran on Virtual HouseCall — an interactive voice counseling platform I designed that anticipated by twenty-five years the personalized health coaching now called digital therapeutics.
Through the late 1990s and 2000s, the methodology reached international scale. Dr. Richard Rubin, my clinical collaborator for a decade, introduced me to Dr. Robert M. Anderson, Ed.D. of the Michigan Diabetes Research and Training Center, co-architect of the empowerment approach to diabetes self-management. Anderson and I worked together for eight years. He wrote of me then as “a kindred spirit” whose work used technology in “a truly humane, empowering, patient-centered way.” That collaboration in turn led Anderson to introduce me to Soren Skovlund at Novo Nordisk's DAWN Programme (Diabetes Attitudes, Wishes and Needs) — the largest psychosocial diabetes survey ever conducted — for which I built the international educational resource deployed across 20+ participating countries.
By 2006, this arc had culminated in a full HIPAA-compliant Diabetes Collaborative Care Platform — a complete operational environment for team-based chronic care. The through-line across all of it: capturing practitioner intelligence and embedding it into systems that scale without losing fidelity. Full detail on this arc — the four-company path, the Diabetes Advisor platform, the DAWN Programme deployment, the 2006 clinical platform — lives on the Diabetes Advisor project page.
The methodology has since extended well beyond diabetes — into mental health relapse prevention, chronic pain management, substance abuse treatment, and cardiovascular care. Diabetes is where the approach was most fully proved out; it is not the boundary of it. By 2017, three decades of this work had crystallized into a coherent platform architecture, set out below.
The vision made visible
By 2017, the methodological commitments threading through three decades of work — semantic ontology as foundation, knowledge engineering as discipline, patient-provider collaboration as the central aim — crystallized into a complete platform conceptualization. Rendered as an integrated four-quadrant architecture: ontologies as the conducting structure that gave the system its organizing principle, AI technologies as the orchestra performing under that direction, analytical capabilities clustered as application services, and a Discover → Learn → Act journey as the user-facing surface.
Click to enlarge
The diagram remains a faithful representation of how CCI thinks about systems that encode clinical wisdom. True personalization, in this view, is not preference settings or user-data records but machine-interpretable, context-aware semantic models — interests, goals, traits, constraints — mapped to evolving domain ontologies. The architecture made this commitment explicit. The applications continue to instantiate it.
The arc that runs through this page has never been solo. Since the earliest days of Health Information Technologies, Dr. Thomas R. Zastowny, PhD has been part of nearly every project on it. A licensed clinical psychologist with a background in computing before he entered psychology, Zastowny came to the intersection of behavior science and information systems from the clinical side. He grasped the vision immediately because he had been trying to build toward it from his own direction. Across more than three decades of collaboration, his role in the work has spanned content development, clinical shaping, informatics collaboration, proposal writing, and program design — the range that a project needs when it requires both clinical instinct and the fluency to translate that instinct into working systems.
Zastowny is a licensed clinical psychologist and health care consultant with 40+ years of applied experience in performance improvement, outcome measurement and assessment, program design and evaluation of health care systems, and behavioral treatment planning. His affiliations include the University of Rochester Medical Center, The Joint Commission (TJC-JCAHO), the Robert Wood Johnson Foundation, CARF, CMS, and NIATx (Network for the Improvement of Addictions Treatment). He has published over 100 scientific articles and book chapters, taught throughout the United States and internationally, and served as a Joint Commission surveyor for more than fifteen years and special projects consultant since 1989. He is the author of A Guide to Performance Improvement in Behavioral Health Care Organizations (1996). International consulting and teaching work has taken him to Italy, France, Ireland, Scotland, Germany, Argentina, Mexico, Korea, Japan, Canada, and the United Kingdom.
His clinical informatics activities include: (1) early work and clinical development in artificial intelligence, (2) development of collaborative and interactive professional networks, (3) design of systems to spread and sustain promising and evidence-based practice, (4) development of multidimensional teaching systems to advance clinical practice, and (5) effective use of regulatory and accreditation systems to improve care and achieve clinical excellence. Each of these threads intersects directly with the work Collaborative Care Intelligence is doing now.
When the work moves into serious infrastructure — cloud architecture, distributed systems, production-scale engineering — it moves into a domain that requires a specific kind of collaborator. David Bradley is that person. A senior systems architect and a trusted colleague of mine for many years, Bradley joins CCI projects when the scope calls for it, bringing 35+ years of designing and implementing high-performance computing infrastructure across financial services, pharmaceutical, telecommunications, printing and imaging, and scientific instruments.
Bradley's career carries a consistent through-line: build systems that hold up when the stakes are real. He was the lead architect on the NYSE Community Financial Cloud, a first-of-its-kind financial cloud for institutions, trading firms, and market-makers — a platform where uptime, security, and performance are not abstractions. As senior systems architect at Nyfix Corporation, he helped grow the company from a $10 million to a $110 million business, scaling its technology infrastructure from a first datacenter to multiple datacenters with 1,000+ systems while designing in-house high-availability solutions that predated commercial equivalents by years. Earlier, at DOTech Corporation, he led an 8-person team as lead architect and developer of Pfizer's Team Repositories — an early document management system for pre-clinical drug documentation that shortened the time from inception to FDA approval. At Sun Microsystems he was named on a Xerox patent and awarded Sun Employee of the Quarter for a 15-month project that saved a $100 million deal by identifying and fixing performance bottlenecks across 160 million lines of client code and Solaris kernel code.
For CCI, Bradley represents the assurance that scale will not become the limit of the work. When investors, partners, or clinical sites raise the natural questions — production deployment at hospital or health-system scale, high availability, secure architecture, integration with existing systems, performance under real-world load — the answer is that a senior architect with more than three decades of production experience at exactly this layer is part of the team. His technical range spans embedded firmware to distributed cloud infrastructure — the full layer stack CCI's methodology needs to become operational. Our collaboration is longstanding and load-tested; when the scope of a project calls Bradley in, the technical foundation is in trusted, capable hands.
Through Collaborative Care Intelligence I am applying the methodology to a next generation of programs. The core commitments have not moved — clinical wisdom encoded into engaging programs, semantic-path personalization, patient-provider collaboration as the substrate of care — but three things have changed.
The audience is direct. Rather than pharmaceutical partnerships underwriting global campaigns, CCI works with independent practitioners who know their patients better than any algorithm — building intelligent programs that respect both.
The tools finally match the ambition. Current AI capabilities enable adaptive coaching conversations, avatar-mediated content delivery, and pattern recognition that were impossible in earlier eras. What used to require months of recording a real clinician's voice can now be built as a warm, responsive presence in a fraction of the time.
The work is being built in partnership with AI — specifically with Anthropic's Claude, not as a tool I use but as a working collaborator. That partnership is what makes the current pace possible. Current projects include the reimagined Diabetes Advisor, the AI-enhanced VIGIL, PainCoach, and the Dynamic Knowledge Architecture book — drafted chapter by chapter through this site.
The current shape of the work: partnerships with independent practitioners on intelligent programs that need more depth, warmth, and methodological rigor than off-the-shelf digital health provides.