I quit medicine to build Superpower’s AI Doctor
This is what I've been up to the last few months...
Earlier this year, I quit my job as a doctor to join a startup called Superpower.
Why?
Well…
It all starts with a seemingly simple question…
“So tell me, why do you want to be a doctor?”
When I was asked that question in my med school interview in the summer of 2017, I thought I knew the answer.
Insert personal story. Insert existential realization. Insert desire to help people mixed with a love of science.
It didn’t feel like I was bullshitting at the time. My answer truly felt real.
But reflecting on that interview 8 years later, the truth was - I didn’t want to study medicine.
When I was 18, I really didn’t know what I wanted to study at all…
Science? Yeah, I enjoyed it.
Helping people was definitely a strong selling point.
The associated money and prestige? Sure, it was nice, but I didn’t really care about that stuff.
My dream was to make a positive, large-scale impact and to leave the world better than how I found it.
Being a doctor seemed like the obvious path. Work in one of the most revered professions. Save lives daily. Make a difference that matters.
But halfway through 2024, I realized that medicine wasn’t what I expected, and it wasn’t helping me achieve my dream.
And so…
I quit.
Do I regret spending 7 years of my life in medicine?
I've spent more than a quarter of my life between med school and the hospital. Quitting after seven years might seem like throwing it all away, but those years taught me more than any other experience ever could.
Yes, there were countless hours of anatomy study, awkward digital rectal exams, and patients screaming about wait times. But there was so much more…
I helped maneuver a baby from a C-section womb in the birth suite, watching him take his very first breaths. A brand new human entering the world.
I gently closed the eyes of patients who had just passed - people who had been with us only hours before - during my palliative care shifts.
I watched ribs pried open to reveal beating hearts during four-hour cardiac surgeries, blood pumping via cardiopulmonary bypass machines.
I held trembling hands as I told a father his CT scan showed brain metastases and his cancer had returned.
I've driven halfway home from work only to turn around and speed back to the hospital because I remembered something critical that I had forgotten to document.
I made the best friends I could ever ask for in med school, people I'll be friends with for life.
I cared deeply for patients and coworkers, and I loved my job.
Despite leaving, I don't regret a second of it all.
So then why did I leave?
In 2024, after around a decade of consistent, nonstop work, I took some time off to re-evaluate my life. I’d been studying since I was 15 to get top scores in high school. I worked throughout medical school to pay rent. I started a company, published research, and worked as a doctor, of course. But I hadn’t really stopped to ask…
Why?
At the end of the day, there was one core desire I had:
My dream was to make a positive, large-scale impact and to leave the world better than how I found it.
And I realized that medicine, while brilliant, wasn’t the answer to this dream.
As a doctor, you work in a healthcare system abounding with problems.
And as a doctor, you’re forced to work within that broken system, but you’re powerless to fix it.
Your impact is strangulated, not by your skills or knowledge or work ethic, but by an environment you’re unable to change.
Whether you're in Australia or America, the healthcare system is crumbling. And doctors - supposedly the driving force of healthcare - are completely powerless to fix it. Most end up starting private practices, jaded and burnt out, held hostage by the sunk costs of their careers.
If I wanted real impact, I realized that seeing patients one by one would never cut it. To make the change I wanted to see, I had to work from outside the system.
Or maybe, rebuild it entirely.
So what exactly is wrong with healthcare?
In first year med school, they painted us a beautiful picture of what healthcare should look like: Spend meaningful time with patients conducting comprehensive histories and exams. Understand them holistically—their emotional health, lifestyle, work stress and social context. Take a preventative approach, solving root causes instead of slapping band-aids on symptoms. Choose surgery or medication only after exhausting other options. Follow up proactively, checking test results and optimizing for long-term outcomes.
Then I started working as a doctor and reality slapped me in the face. Patient consultations became six-minute speed rounds interrupted by pagers, phone calls, and competing priorities. Managing conditions meant finding the fastest route to discharge—surgery or drugs, whatever worked quickest. Preventative health? Lifestyle changes? No time for that luxury.
Proactive care got dismissed as "infeasible." We delivered reactive care whenever patients showed up, then dumped the rest on overwhelmed primary care providers who faced the exact same constraints.
The medicine I'd idealistically envisioned had warped into sick-care, not healthcare.
This broken system revealed three fundamental constraints that corrupted everything:
Time
Money
Volume1
There's never enough time to deliver quality care. Patients can't afford higher quality options. There are simply too many patients and not enough providers. Meanwhile, healthcare staff are overworked, underpaid, and drowning in endless patient lists.
Traditional medicine isn't built around optimal care for individuals. It's built around these three constraints. And now the whole system is at breaking point.
But how do we fix it?
Save your drum roll because you’ve already guessed the answer.
Artificial Intelligence
Before you groan at yet another "AI will save us all" pitch, hear me out.
Currently, doctors are the decision-makers of healthcare.
But, while they are a crucial part of the system, they take years to train, they are expensive, and they can only see one patient at a time.2
Time.
Money.
Volume.
And that's assuming all doctors deliver equivalent care quality (spoiler: they don't)
AI changes everything. It enables scalability and quality control of what's currently an expensive, service-based model. As AI models improve, we'll create AI doctors that outperform human doctors across every domain and specialization.
We can deliver ideal healthcare at scale by training AI with medical best practices, completely ignoring the constraints of time, money, and volume that cripple human doctors. But here's where it gets interesting. We can't just build an AI that mimics human doctors—that's the horseless carriage problem. We need to build an entirely new healthcare experience from the ground up.
One of Superpower's founders, Max, poses this thought experiment: Imagine finding the world's best doctor and offering them $100 million to focus on just one patient for an entire year. No on-call shifts, no other distractions.
That doctor would read through the patient's notes daily. Consult specialists constantly. Stay current on bleeding-edge research. Explore novel treatments from obscure case studies. Schedule healthy meals and exercise sessions. Manage medications with obsessive precision while monitoring for side effects. Understand the patient psychologically and emotionally. Reach out before problems develop, checking vitals and lab reports proactively.
Now imagine scaling that level of meticulous care to everyone, everywhere.
That's the healthcare experience we're building. No constraints. A completely new paradigm.
Preventative, personalized, and performance enhancing. Not just for those who can afford it, but for all.
But AI Doctor how?
I know what you’re thinking:
“Slapping AI on any and all problems doesn’t solve them.”
You’re right, it’s easier said than done.
But it’s achievable.
To understand why, we first need to understand what doctors actually do.3
At its essence, a doctor’s role is the integration and processing of three datasets.
Patient Data (symptoms, medical history, lab results, imaging, genetics, lifestyle habits)
Medical Knowledge (research papers, clinical trials, treatment guidelines, diagnostic frameworks)
Management options (medications, therapies, procedures, counseling, lifestyle interventions)
Doctors act as the processor between these three datasets. They:
Ingest a patient’s data relevant to the clinical scenario, including reading previous reports and eliciting medical history to understand the full diagnostic odyssey
Overlay this data on the current body of medical knowledge to draw connections, make predictions, and determine the best course of action
Output a set of recommendations in the form of a care plan that encompasses relevant management options
But being a doctor is difficult and has a high academic barrier because the three aforementioned datasets present an increasingly large cognitive workload:
Approximately 30% of the world’s data volume is being generated by the healthcare industry. By 2025, the compound annual growth rate of data for healthcare will reach 36%.
Medical knowledge used to double every 50 years. Now, a new PubMed article is published every three minutes, and medical knowledge doubles every 73 days. It’s impossible to keep up.
Despite new treatment options and guidelines being published annually, it still takes 17 years for evidence to influence how doctors practice medicine in a clinical setting.
Historically, medicine dealt with information overload through specialization. Renaissance doctors could master all medical knowledge because there wasn't much of it. As knowledge expanded, we created specialties—cardiology, nephrology, pharmacology. Now we have sub-specialists: not just orthopedic surgeons, but orthopedic shoulder surgeons.
We're reaching the limits of human specialization. The next evolutionary step is AI doctors that can store and synthesize petabytes of health data to achieve optimal outcomes.
But AI Doctor why?
Superpower is making some bets about consumer behavior that we have a lot of conviction in. These bets make the AI doctor a product that needs to exist:
Healthcare access is broken. Patients want fast, convenient care available anywhere, anytime. They're already going to "Doctor Google" before seeing real doctors.
Healthcare is too expensive. Americans are avoiding medical care due to cost at record rates.
Trust in healthcare is eroding. Consumer distrust is growing across all generations.
Thus, we believe that to overhaul the healthcare system, we need to:
Make healthcare convenient
Make healthcare cheap (or free)
Make healthcare high-quality and trustworthy
The AI doctor will be able to fulfill these consumer desires. Rather than googling symptoms before going to their PCP, consumers will go straight to their AI doctor to solve all their health issues.
In doing this, our AI Doctor becomes the “front door” of healthcare by addressing consumer desires for trusted, cheap, and accessible healthcare.
Will people trust the AI doctor?
One of the largest obstacles we’ll need to overcome is trust.
I believe that trust in the AI doctor comes down to five key factors:
Accuracy - we need to use evidence-based best practices with clear and transparent reasoning
Naturalism - we need the AI doctor to feel like a real person, empathetic but not sycophantic, almost being able to cross the uncanny valley
Boundaries - The AI must know when not to act and when to escalate to humans. Overconfidence kills trust instantly.
Human oversight - Human doctors must review diagnoses and treatment plans, especially for complex cases.
Interpretability - We need to understand how the AI reaches conclusions, ensuring its reasoning maps to sound clinical logic
Initially, we're focused on getting the core logic right—clinical utility and safety come first.
Soon, focus will shift to the “feel” of an interaction with the AI doctor. Medicine and health are sensitive topics for most, and the taste of an interaction is almost as important as the outcome. Tone, persona, and feel of an interaction are important, but malleable. After the core logic is accurate and high quality, focus will switch to the behavioral characteristics of the AI doctor to build trust and determine how it’ll communicate things like uncertainty, empathy, or urgency.
Paradoxically, AI enables both standardized and personalized care. Standardized because AI follows best practices without human error or clinical bias. Personalized because AI can analyze vastly more patient data than any human doctor.
This is key to building user loyalty, establishing clinical confidence and encouraging patients to speak freely and return for repeat consultations.
In an AI Doctor world, what happens to human doctors?
Will I be putting doctors (and some of my best friends!) out of business if Superpower achieves this goal?
No.
I know that many of the AI products being released cause widespread fears of job risk. There’s also a lot of fear-mongering about this predicted AI future. AI will inevitably change our economy and humanity’s future; that’s undeniable.
To be clear, I don’t feel bad about the AI products we’re building, though. In fact, I feel quite positive.
In my opinion, AI is being used across industries to do one of two things:
To optimize efficiency
To improve access
The first replaces humans to cut costs, like AI customer service agents that encourage companies to fire human representatives. The second improves access to things that are difficult or expensive. People who might otherwise be unable to access high-quality coaches, lawyers, or doctors will soon find these services democratized thanks to AI. Superpower is firmly in this access camp. We're not trying to optimize away doctors' jobs - we're trying to give healthcare access to people who couldn't afford quality care before.
Even with widespread AI adoption, hospitals will still need human doctors. Emergency medicine, surgery, and complex cases will require human expertise (for the time being), and I don’t see this changing anytime soon.4
What about the humanity of healthcare?
A common platitude I’ve heard is that medicine is an inherently and unavoidably human profession that will never be able to circumvent a human doctor driving encounters.
Funnily enough, I don’t think it’s the doctors claiming this.
Doctors are socially revered, but in my opinion, that’s more a matter of branding than anything else. Having worked with doctors and as a doctor myself, they aren’t the ones who spend the most time with patients.
Doctors are the ones who make decisions, but they don’t have a monopoly on the humanity of healthcare. The humanity of healthcare can come from nurses, therapists, and other caregivers.
And as we’re seeing with today’s patients, a future generation more mobile and globalized than ever, they simply don’t care. They need access that’s convenient and cheap, more than human-based and service-led.
Today's patients—especially younger, jet-setting generations - prioritize convenience and affordability over traditional bedside manner. Medicine has a history of resisting technological change, but consumer behavior can't be controlled.
The world is changing at an increasing speed. Healthcare needs to catch up to the future, not recreate the past.
Aren’t you only fixing part of the problem?
This is the question I think a lot of my doctor friends will ask.
Does this work actually achieve my dream of making a large-scale positive impact?
Increasing access doesn't solve everything - I’ve seen it with my own eyes. Even Australia's free public healthcare system struggles with the same fundamental issues.
Ask any doctor and you’ll learn that the most difficult, time-consuming, and resource-draining patients are the ones who don’t engage with the healthcare system at all. They’re the ones who need the most help, but are also the hardest to help.
Why bother building an AI doctor if we can’t help the people who really need healthcare the most?
As someone who has worked in public hospitals, I can understand this sentiment firsthand. However, I don’t think what we’re doing is futile. Superpower’s goal is to change the entire healthcare system, but there may be some parts of the healthcare system we’re unable to make a dent in - and that’s okay.
Even if we initially serve only the socio-economic and health-conscious top 10% who engage regularly with the healthcare system, that's still millions of people getting better care, millions in value generated, and millions of healthcare spending dollars saved. Superpower's goal is to scale this to billions and to scale globally.
But ultimately, the biggest unlock with accessibility is engagement and thus, prevention.
Heart disease, diabetes, and other preventable chronic conditions account for the vast majority of healthcare spending and system burden. If we can make even a small dent in their prevention through increased engagement, we'll have a massive impact, no matter the population demographic we serve.
Goodbye, medicine, for now…
When I graduated from medical school and signed the Hippocratic Oath, I pledged to do no harm, practice with integrity, and act in patients' best interests.
While I may have left the industry, I still take that oath very seriously.
Leaving medicine was the hardest decision I've ever made. I truly loved being a doctor.
But, my dream is to make a positive and large-scale impact, and to leave the world better than how I found it.
I just knew medicine wouldn’t help me achieve that ambition.
When people ask me how I feel about leaving medicine, my answer is “medicine was a girlfriend but not a wife.”
I still love medicine dearly, and now that I’m gone, I often miss it. It was definitely a love-hate relationship, maybe a little toxic at times, but one that I valued highly. I was good at it, and I worked extremely hard.
Who knows. Maybe once Superpower has distributed healthcare around the world, one day I’ll happily reunite with my old love and go back to being a junior resident at a hospital somewhere in a small rural Aussie town.
But for now, I know what my mission is—what Superpower's mission is—and that's what I'm married to.
World’s best healthcare.
For 1 billion people.
For free.
Stay tuned 🧡
One could say volume is the precursor constraint for time. However, I want to treat these metrics differently as individual patients can take a long time to evaluate and treat thoroughly, making this a subtly separate issue from the sheer volume of patients seeking care.
The simple solution people often propose is to train and hire more doctors. This would solve the volume and time issues as we’d increase the ratio of doctors per patient. However, we’d still encounter the money issue, furthered by the fact that these new doctors would need to be paid for by someone. Governments? Insurers? Patients? Nobody wants to foot the bill (And don’t even get me started on training program bottlenecks creating artificial supply/demand constraints thanks to the incentives medical colleges have to remain ultra-exclusive)
This is a non-technical explanation of the concept of an AI Doctor, written mostly for friends and family as a career update. If there’s interest (and company intellectual property permitting) I may eventually write a more technical exploration of our approach to building an AI Doctor. Realistically, this will probably be a retrospective piece after the product is released
In the near-term future, regulatory compliance is also a big contributor to my opinions here. Human doctors will continue to be the ultimate decision makes, but their decisions will simply be supported by agentic work.