A surgeon's work is a chain of decisions disguised as a pair of hands. Before an incision comes the history, the scan, the risk calculation, the consent, the plan and the plan for when the plan goes wrong. Adam Oskowitz spent years inside that chain. At UCLA he trained in general surgery, served as chief resident, and then specialized in vascular surgery. At the University of California, San Francisco, he became a clinician, researcher and teacher. By 2025, his title at Zuckerberg San Francisco General had acquired another word: chief.
Then came a second system with its own chain of decisions. Doctronic, the company Oskowitz co-founded with technologist Matt Pavelle, begins with a conversation on a screen. It gathers context, produces a structured assessment and can pass that record to a licensed physician. The interface is simple because the infrastructure behind it is not. The company describes specialized agents, safety checks, clinical escalation and a medical practice licensed across the country. Oskowitz's role is to make sure the product does not confuse a fluent answer with a finished job.
This is the useful way to understand his path. He did not wake up bored with surgery and decide that software looked fashionable. His clinical, scientific and startup careers are all versions of the same problem: how do you intervene in a complex system, at the right moment, without losing sight of the consequences?
The apprenticeship in consequences
Oskowitz's résumé has the compression of someone who stayed in school long enough to collect several professional languages. He earned a master's in public health at Dartmouth, then an MD and PhD at Tulane. He entered general surgery at UCLA in 2010, completed residency in 2014 and finished a vascular fellowship in 2016. The American Board of Surgery certified him in general surgery in 2015 and vascular surgery in 2018.
Each credential enlarged the frame. Public health asks what happens across a population. A PhD asks whether a claim survives experiment. Surgery asks whether judgment holds up in the room. At UCSF, those frames converged. Oskowitz treated complicated vascular cases while studying whether stem cells and immune cells could be directed toward damaged aortic tissue. The work was not about making an existing procedure a little faster. It was about changing the available intervention.
In 2020, that work earned a $400,000 R21 Trailblazer Award from the National Institute of Biomedical Imaging and Bioengineering. The three-year grant backed an ambitious approach: engineer regulatory T cells to travel toward injured aortic tissue and reduce the inflammation associated with aneurysm growth. Oskowitz described the goal plainly - redirect a patient's own cells toward injury and help the body repair itself. Three years later, UCSF funded another project on the role of T cells in aortic aneurysms.
The details are scientific, but the founder instinct is recognizable. Start with a system that already contains specialized actors. Give them better information. Direct them toward the place where a decision matters. Measure what follows. The unit later changed from a cell to an AI agent, but the taste for coordinated systems remained.
A company built at the handoff
Doctronic began with a first-principles question shared by Oskowitz and Pavelle: what would a care system look like if it were organized around the patient rather than the institution? The two founders brought opposite halves of a useful argument. Pavelle had built consumer technology. Oskowitz knew what happens when information crosses a clinical boundary badly.
Their answer was not to hide the boundary. A person can start with the AI, free and available at any hour. The conversation becomes a summary and doctor's note. If the case needs a person, the context moves with it to one of Doctronic's employed physicians. A handoff that once meant retelling the same story can become a continuation.
“There's probably a hundred times more care that could occur if there's zero friction in the system.”Adam Oskowitz, speaking after Doctronic's Series B
Oskowitz likes to explain the idea through the small questions. In a March 2026 interview on TBPN, he described users who return repeatedly, sometimes asking what to eat for breakfast in light of a lab result. These are not dramatic encounters. That is the point. A conventional appointment creates a minimum viable concern: the question must feel important enough to schedule, wait, travel, pay and ask. An always-open conversation removes that threshold.
His estimate that zero friction could produce one hundred times more care is part provocation, part product thesis. Friction does more than slow a transaction. It suppresses demand. Remove it, and behavior changes. People ask earlier, more often and with less ceremony. The system must then distinguish between a useful answer, a necessary visit and an urgent escalation. Access creates its own sorting problem.
A narrow door to a larger room
One of the company's consequential moves began narrowly. Under Utah's AI Learning Lab regulatory sandbox, Doctronic started renewing a defined set of existing prescriptions without a physician acting as the intermediary. Oskowitz called renewals “a safe place to land.” That phrase contains a clinician's approach to product expansion: choose a bounded decision, enumerate the risks, create automatic escalation, and publish the result.
Doctronic says its system performs dozens of safety checks during a renewal. The company also reported 99.2 percent alignment between its treatment plans and those of board-certified physicians in a 500-patient evaluation. Oskowitz talks about these figures the way an operator talks about a production system. Agreement, throughput, retention and escalation are not abstract benchmarks. They tell him whether people trust the loop enough to return.
- Completes medical and doctoral training at Tulane.
- Finishes vascular fellowship at UCLA and moves into UCSF surgery.
- Receives a $400,000 NIH Trailblazer Award for cell-based vascular research.
- Co-founds Doctronic with Matt Pavelle.
- Is named Chief of Vascular Surgery at Zuckerberg San Francisco General.
- Doctronic announces a $40 million Series B and more than $65 million raised.
Capital followed the operational case. Union Square Ventures invested at seed. Lightspeed led a $20 million Series A in September 2025. Six months later, Abstract and Lightspeed co-led a $40 million Series B, bringing total funding above $65 million. The company said it had grown fifteenfold after the Series A, reached eight-figure annualized revenue and attracted more than 300,000 unique weekly visitors.
Those numbers make an efficient fundraising slide. Oskowitz, however, singled out a quieter one: repeat physician visits had nearly tripled. “Patients aren't just trying us,” he said. “They're coming back.” Novelty can produce a spike. Return behavior suggests the product has found a place in someone's routine.
What builders can take
The fashionable reading of Oskowitz's story is that domain experts should start AI companies. The more precise lesson is that expertise matters when it changes the architecture. His clinical background appears in the unglamorous parts: structured intake, multiple checks, clear escalation, a record that follows the user, and attention to who owns the outcome.
Define the consequential action before choosing the interface or model.
A handoff should continue the conversation, not force the user to restart it.
Escalation is core product behavior, especially when confidence has limits.
Repeat use says more about utility than a rush of first-time curiosity.
There is also a lesson in how Oskowitz handles not knowing. During the TBPN appearance, the hosts asked him to weigh in on a controversial peptide. He did not reach for an impressive-sounding answer. He said he did not know what it was and deferred to the system. The moment was brief, almost comic, and more revealing than a polished pitch. In a field rewarded for fluent certainty, the surgeon-founder left uncertainty visible.
That restraint sits beside considerable ambition. Oskowitz wants the Utah model to travel to other states and wants the system to take on more work over time. He has argued that a shortage of clinicians is fundamentally a human-capital problem. If software can safely handle a larger share of routine decisions, he says, part of that constraint becomes a compute problem.
The operative word is “becomes.” Software does not erase the human system around it. It changes where people enter, which work reaches a physician, and how much context arrives with it. Oskowitz still occupies both systems: the hospital division where expertise remains scarce and physical, and the company where he is trying to make parts of that expertise travel.
His career has moved from hands, to cells, to teams, to agents. At each scale, the challenge is coordination. Get the right information to the right actor. Make the next action legible. Build a route for the exception. Accept that intervention creates responsibility.
A good decision loop is not spectacular. It is available when a small question appears, careful when the answer carries weight, and honest about the moment it needs help. That is the system Oskowitz has spent his working life learning to build - first one patient at a time, and now with no waiting-room walls.