Early detection can make a life-changing difference in cancer care, but access to timely screening and specialist expertise remains deeply uneven. In many communities, especially those far from major medical centers, patients face delays not because treatment does not exist, but because the health care system lacks the time, workforce, and infrastructure to reach them early enough.
That challenge is what led Mario Minor Murakami Junior, a postdoc at Harvard Medical School, to create AI Pathology, a venture that uses artificial intelligence to support earlier cancer detection and strengthen care delivery in underserved settings and received SIFF funding in 2025. Drawing on experience in medicine, computer science, and imaging research, the AI Pathology team is building tools designed to help clinicians and frontline health workers identify concerning skin lesions sooner and make more informed decisions.
In a recent conversation with the Harvard Innovation Labs, Murakami shares what inspired him to start AI Pathology, the responsibility that comes with building in health care, and the unexpected moments that shaped the company’s journey.
What moved you to start your venture?
One of the most defining moments of my life happened during my final year of medical school, at the start of the COVID-19 pandemic. I was volunteering as an emergency trainee in São Paulo, Brazil. Because there were not enough doctors, medical interns were asked to help in overcrowded emergency rooms in underserved neighborhoods.
One night, standing in an improvised emergency room with my professors, I realized how limited our help really was. Even with experienced doctors working nonstop, we could only treat a small portion of the patients arriving each hour. I saw people get worse and die, not because their illnesses could not be treated, but because there simply were not enough people, time, or resources. That feeling of helplessness stayed with me, especially knowing we were serving communities that had long been overlooked.
At the same time, I was reading “Deep Learning” which explains how artificial intelligence can enable data-driven decisions . The contrast was hard to ignore. We were living in a time of major technological progress, yet right in front of me people were losing their lives due to system failures that technology could help fix. I felt frustrated and motivated, realizing that many powerful tools were being used for social media or low-impact problems, while health care was falling behind.
That experience changed my direction. I understood that as one doctor, my impact would always be limited by time and place. But by combining my background in computer science with medicine, I could build tools that extend care and reach far more people. That was the moment I decided to become a health-tech entrepreneur. AI Pathology grew out of that decision and reflects my belief that artificial intelligence should be used to serve people and save lives.
What impact are you hoping to achieve?
My vision is to make early cancer detection as accessible and routine as checking your blood pressure.
Specifically, I want to eliminate the gap between the first visible signs of disease and medical decisions, especially for people who live far from specialists. In practical terms, this means that a rural worker in Brazil, a primary care clinic in the public health system, or a community health agent with only a smartphone can detect high-risk skin lesions early enough to change outcomes.
In 10 years, success for me looks like this: millions of people screened annually through AI-powered tools embedded directly into primary care and public health workflows; skin cancers detected months or years earlier than they otherwise would have been; and specialist time reserved for the patients who truly need it. AI Pathology is not meant to replace doctors; it is meant to empower them.
What’s the hardest part about building?
The hardest part has been learning how to build trust while carrying responsibility at scale.
Early on, I realized that in health care, you cannot afford to “move fast and break things.” This became very clear during our first pro bono screening project in a rural community in Goiás. We had a Convolutional Neural Network (CNN) model that performed very well in testing, but I still found myself lying awake at night thinking about edge cases, about the one patient who might be classified incorrectly and what that could mean for their life. The hardest part is not the technology; it is carrying the responsibility of knowing that your decisions affect real people, often in vulnerable situations, and still having the courage to move forward.
Have you always been involved in this work?
I transitioned, but looking back, everything I did was preparing me for this.
I was trained as a physician, but before medicine I had a strong background in computer science and research in MRI imaging. For a long time, those paths felt separate. The transition happened when I realized that the limitations I felt as a doctor—like time, geography, and workforce shortages—were exactly the problems technology is best at solving. AI Pathology was born at the intersection of those two paths, when I stopped choosing between them and decided to merge them.
What’s the most unexpected moment you’ve had building your venture?
The most unexpected moment in this journey was how I met my co-founder.
At the time, I was being interviewed on a podcast a project raising funds to send Brazilian medical interns to Harvard laboratories. Without my knowing, another guest was being interviewed separately on the same podcast. That guest was Will Boelcke.
After the episodes were released, we each watched the other’s interview. It was one of those rare moments when you immediately feel a connection, even through a screen. Will spoke about losing his father to skin cancer and how that experience motivated him to work on something that truly mattered. At the same time, I was working in medical imaging research, trying to turn science and AI into something that could make a real difference in people’s lives. When Will and I finally spoke, the conversation did not feel like a first meeting; it felt like continuing a discussion that had already begun. His personal motivation and my clinical and technical background fit together naturally. Looking back, it still feels surprising that a podcast interview meant for a completely different purpose ended up connecting the two people who would later start AI Pathology. But that moment is what started everything.
What’s been one of the coolest moments?
We reached nearly 3,000 screenings and close to 100 diagnoses through pro bono work before we were even a “company” in the traditional sense. There was no formal branding, no sales team—just technology being used where access to care was limited. As a founder, it is rare to see validation that early and that directly. Those results made it clear that this was not just an interesting idea; it was something people truly needed.
Any lessons learned?
Don’t confuse intelligence or credentials with being ready.
I spent years training as a physician and researcher, and I still underestimated how different it is to build a company. Success does not come from being the smartest person in the room. It comes from listening, moving quickly, being honest about what you do not know, and building the right team early. If I could save future founders from one mistake, it would be trying to do everything alone for too long.
