Dr. Sharad Maheshwari MD - Narrative Biography & CV

on
From Images to Judgment | Biography

From Images to Judgment 🧠

Lessons from Three Decades of Medicine, Technology, and Responsibility

When people hear that I work in artificial intelligence governance today, they often assume that my journey began with AI.

It did not.

In many ways, it began with a film hanging on a light box. 🎞️

When I entered radiology in the 1990s, medicine occupied a very different world. Images were printed on films. Reports were dictated onto tapes. Information moved slowly. A diagnosis depended not only on knowledge but also on patience, observation, and experience.

At the time, I could not have imagined that I would one day be discussing generative AI, algorithmic accountability, automation bias, or governance frameworks. Yet the questions that occupy me today are rooted in lessons I began learning long before artificial intelligence entered healthcare.

Looking back, I realize that my professional life was shaped by four parallel journeys: a journey through technology, a journey through teaching, a journey through clinical medicine, and ultimately a journey into governance.

For many years, I did not realize they were all leading to the same destination. 🎯

🔬

The Technology Journey

My fascination with technology began early.

At Jaslok Hospital in Mumbai 🇮🇳, I had the opportunity to work with advanced MRI systems during a period when imaging technology was evolving rapidly. For a young radiologist, MRI felt almost magical. Structures that had once been invisible could now be visualized with extraordinary detail. Every new sequence seemed to reveal another layer of anatomy and pathology.

Crucially, I was mentored by Dr. Srinivas Desai, who provided an incredible environment for academic exploration. He gave me the intellectual freedom to truly master the equipment. We would dedicate time—often staying late after routine clinical lists were safely completed—to optimize protocols, test new sequences, and deeply understand the physics behind the images. This rigorous, hands-on learning taught me early on that technology must be deeply understood to be used responsibly.

Technology was expanding the boundaries of what physicians could see.

I was captivated.

Yet what fascinated me was never technology itself.

It was what technology allowed human beings to do. 💡

Years later, a scholarship opportunity took me to Ann Arbor, Michigan. The University of Michigan exposed me to a different dimension of medicine. I encountered an academic culture where excellence was not measured solely by clinical expertise but also by the ability to teach, mentor, and inspire.

Among the people who left a lasting impression on me was Dr. Suresh Mukherji. From him and many others, I learned that expertise becomes meaningful only when it can be shared.

Knowledge that remains with one individual eventually disappears.

Knowledge that is taught creates generations. 🌱

That lesson would influence everything that followed.

Soon afterward, I moved to Canada 🇨🇦 for advanced training.

The University of Toronto was a revelation.

What I encountered there was not simply excellent radiology. It was a healthcare ecosystem increasingly shaped by digital technology, informatics, collaboration, and systems thinking. Across Toronto General Hospital, Princess Margaret Hospital, Toronto Western Hospital, and the wider academic network, I saw medicine becoming interconnected in ways that were impossible in the film era.

It was also here that I was influenced by Dr. Stephanie Wilson—a brilliant sonologist, educator, and researcher. Her relentless drive to try new things and pioneer advanced ultrasound techniques showed me how clinical curiosity should dictate technological progress, not the other way around. She embodied the spirit of a true clinician-scientist.

Technology was no longer merely producing images.

Technology was beginning to reshape healthcare itself.

The experience changed how I thought about innovation.

Over three decades, I witnessed nearly every major technological shift in diagnostic imaging firsthand:

  • → Film to PACS
  • → Analog to digital workflows
  • → Local reporting to teleradiology
  • → Computer-aided detection to deep learning
  • → Deep learning to generative AI

Technology changes. Responsibility does not. Every report still carried one signature. 🖋️

At McGill University in Montreal, another lesson emerged.

Working with outstanding mentors, including Dr. Caroline Reinhold, reinforced a principle that has remained with me throughout my career.

The best clinicians embrace innovation without surrendering judgment.
Technology can enhance expertise.
It cannot replace wisdom. 🦉

At the time, I viewed these experiences as separate chapters.

Only much later did I realize they were teaching the same lesson from different angles.

📚

The Teaching Journey

When I look back on my mentors, I notice something that escaped me at the time.

The people who influenced me most were not simply skilled radiologists.

They were educators. 👨‍🏫

Whether in Mumbai, Ann Arbor, Toronto, Montreal, or later in my own practice, the individuals I admired shared a common characteristic. They were committed to helping others think more clearly.

Teaching is often misunderstood as the transfer of information.

It is not.

Teaching is the cultivation of judgment. ✨

A teacher's task is not to provide answers.

It is to help people ask better questions.

The seeds of this dedication to teaching were planted early. During my MD residency, I would make personal visits to KEM Hospital in Mumbai just to study the legendary teaching files curated by Dr. Ravi Ramakant. Seeing how meticulously he compiled cases to educate the next generation was profoundly inspiring. That early encounter left an indelible mark—it is the direct reason why, today, I maintain my own curated teaching archive of over 200,000 abdominal imaging cases. 🗂️

Years later, our paths crossed again when Dr. Ramakant served as the director at Kokilaben Hospital. Beyond academics, his focus had a profound philosophical core: empathy. He constantly reminded us that no matter how advanced our diagnostic tools become, the application of technology must always remain profoundly human. 🤝

Over the years, teaching became an increasingly important part of my professional identity. Residents, fellows, lectures, conferences, articles, courses, blogs, podcasts, teaching archives, and eventually AI literacy initiatives all emerged from the same impulse.

I wanted to help people think.

That aspiration would later become central to my work in governance.

Because governance, at its core, is also educational.

It is society's attempt to teach itself how to use power responsibly. 🏛️

🩺

The Clinical Journey

Technology and teaching were important.

But clinical medicine remained the foundation.

Since joining Kokilaben Dhirubhai Ambani Hospital in 2008 🏥, I have spent thousands of hours interpreting scans, participating in multidisciplinary discussions, supporting liver transplant programs, collaborating with oncologists and surgeons, and helping make decisions that affect real patients and real families.

This is where theory encounters reality.

A textbook rarely describes uncertainty the way clinical practice does.

A conference presentation rarely captures the consequences of a missed diagnosis.

Medicine is not pattern recognition alone. A scan never speaks for itself.

Images gain meaning only when interpreted within the context of a patient's history, symptoms, risks, and future consequences. This appreciation for context would later become central to my thinking about artificial intelligence.

There is an epistemological divide between probabilistic systems and deterministic responsibility.

An algorithm may generate a probability. A physician must make a definitive decision. 🤝

Over time, I came to understand that medicine is not primarily a discipline of answers.

It is a discipline of judgment.

Radiologists live in a world of probabilities.

Every image contains ambiguity.

Every diagnosis contains uncertainty.

Every report represents an attempt to answer a deeper question:

  • What do we know? 🔍
  • How confidently do we know it? 📊
  • And what should we do next? 🛤️

Without realizing it, I had spent decades practicing applied epistemology.

⚖️

The Governance Journey

When artificial intelligence arrived in healthcare 🤖, I did not see it as an entirely new problem.

I saw it as an old problem in a new form.

AI raised the same questions clinicians confront every day.

What is the evidence? 📉
How reliable is it? 🛡️
What are its limitations? 🚧
When should it be trusted? ✅
When should it be challenged? 🛑

The more I explored AI, the more I became interested not in its intelligence but in its consequences.

As AI development accelerated, one overarching concern stood out above all others:

The Ethics-to-Execution Gap. ⚠️

Coding had become incredibly accessible. For a student or individual builder, writing an algorithm was easier than ever. But in this rush to create, stopping to consider ethics for every model was rare.

In medicine, we operate differently. Every study, protocol, and intervention must go through rigorous review. As a member of the Institutional Ethics Committee at Kokilaben Hospital, I am constantly reminded that patient safety and ethics cannot be afterthoughts.

This contrast led me to found BeResponsibleAI. Initially, I envisioned it as a "virtual ethics committee for individual coders"—a way to inject clinical guardrails into the building process.

But as my journey continued, my understanding of the problem expanded. The danger of AI was not just in its creation, but in its daily clinical use. Concepts like automation bias—the gradual erosion of human vigilance—showed me that an algorithm does not merely participate in the decision; it participates in the decision-maker.

I realized that governance could not just be a checklist at the beginning.
 It had to be an end-to-end architecture: from thinking, to coding, to deployment, and enforcing runtime accountability until the very end of the system's life. 🔄

Medicine has always understood something that technology sometimes forgets.

Not everything that can be done should be done.

Increasingly, I found myself asking a question that seemed absent from many conversations about innovation:
Should this system exist at all? 🧐

Furthermore, I observed a significant AI literacy gap in healthcare. Clinicians needed practical tools to evaluate these lifecycle risks. To bridge this, I co-authored the Clinical AI Made Easy for Medical Professionals curriculum—a comprehensive program designed specifically to help doctors, nurses, and medical students understand and safely integrate AI.

The central challenge of clinical AI is reconciling probabilistic computation with deterministic accountability structures. This thinking led to the development of a layered governance architecture—three complementary frameworks that address different stages of the AI lifecycle:

🧑‍⚕️

PCCM (Patient-Centered Concentric Governance Model)

Establishing Patient-Centered Legitimacy

PCCM asks why AI should be governed. It is a three-layer accountability framework that places patient safety at the absolute center. The middle layer distributes stakeholder accountability (reminding us that clinicians retain professional and medico-legal responsibility), and the outer layer consists of the responsible AI infrastructure required to operationalize that safety.

🏛️

PRIME (Pre-Governance Intelligence & Risk Evaluation)

Pre-Development Admissibility

Governance must begin at ideation, before a single line of code is written. PRIME shifts the discussion from "Can we build it?" to "Should we build it?" It serves as a structured gatekeeper, evaluating the purpose, societal desirability, workflow integration, and ethical acceptability of a proposed system to prevent technically accurate but clinically misaligned systems from being developed.

🛡️

RATSe-Health (Resilient AI Trust Score)

Operational Governability

RATSe bridges the "ethics-to-execution" gap by addressing the operational question: Can this deployed system remain safely governable over time? It structures runtime accountability across six dimensions (Responsible, Accountable, Transparent, Safe, Ethical, Sustainable). It ensures mechanisms exist to detect dataset shift or operational drift, providing explicit clinician override capabilities to preserve human authority against automation bias.

AI can advise. Humans must judge. ⚖️

Convergence 🌐

Today, people sometimes ask whether I am a radiologist, an educator, an AI builder, or a governance advocate.

The answer is that I never consciously chose between those identities.

They grew together. 🌳

  • The technology journey taught me what was possible. 🚀
  • The teaching journey taught me how knowledge spreads. 📖
  • The clinical journey taught me the weight of responsibility. ⚖️
  • The governance journey taught me how those lessons could be applied to emerging technologies. 🧭

"What began with images ultimately became a search for judgment." 👁️

The Mission Evolves

From reading scans to designing decision systems.

From diagnosing disease to governing intelligence.

Across India, the United States, and Canada, across hospitals, universities, conferences, classrooms, and clinics, one question remained remarkably consistent:

How do we use increasingly powerful technologies without losing human judgment? 🤔

That question guided me when MRI transformed imaging.

It guided me when digital systems transformed healthcare.

It guided me when AI entered medicine.

And it continues to guide me today. ✨

Because technology will continue to change.

The responsibility to use it wisely will not. 🛡️

© 2026 Biography. Designed with care.

Comments

Post a Comment