Elad Walach Learned That the Hard Part of AI Comes After the Algorithm

The first clue is the support line. In Aidoc's early years, when its artificial intelligence was beginning to appear inside hospitals, the company's chief executive and chief technology officer answered customer calls themselves. There were grander jobs available. Elad Walach could discuss neural networks, regulation, the future of diagnosis or the appetites of venture capital. Instead, the co-founder was close enough to hear what happened when a promising model met a crowded workday.

This is the useful tension in Walach's career. He has moved fast for most of his life. He entered university at 15. He learned computer science and physics, graduated from Israel's Talpiot technology program and spent the next six years developing algorithms for the air force. He listens to audiobooks at 2.5 times speed. Yet the company he has led since 2016 operates in a world where speed must continually introduce itself to caution.

Walach's subject is clinical AI, but his real preoccupation is adoption. A clever model in a laboratory can win applause. A useful system in a hospital must win something less theatrical and much harder: a place in the workflow. It must arrive at the right moment, make sense to the people doing the work, survive scrutiny and show that something improved. Intelligence is welcome. Evidence gets a chair.

The boy who kept taking things apart

Walach was born in Haifa and has described himself as a typical high-school science nerd. His father was a scientist at IBM and one of the people who pushed the company toward healthcare imaging. Curiosity became a family dialect. When the younger Walach read about an MIT Media Lab project using a webcam to measure pulse, he built his own version. He was not yet chasing a market. He wanted to know how the trick worked.

Talpiot changed the scale of the puzzle. The selective military program combined science with leadership and placed Walach among people who would later become his co-founders. He has written that he arrived strong in science and conspicuously short on leadership skills. The program taught him that invention is usually plural: teams, coordination, deadlines and responsibility. The lesson sounds sensible on paper. Under pressure, it becomes muscle memory.

It also gave Walach a personal clock for management. He has described the need to surface from daily work, scan the horizon and think ahead before circumstances do the thinking for you. That instinct later found a civilian counterpart in Kaizen. One lesson came from a military program designed to prevent technological surprise; the other came from a philosophy of small, regular improvements. Together they make an unusual executive temperament: impatient about the destination, methodical about the route.

After nine years in the military technology system, including leading AI research, Walach, Michael Braginsky and Guy Reiner wanted to build something together. Their technical credentials were impressive. Their knowledge of hospitals was not. Walach has been blunt about the deficit: they knew little about business, little about healthcare and even less about American healthcare.

So they went looking rather than pretending. For roughly a year, they spent time in hospitals including Sheba Medical Center, watching and asking. A recurring picture emerged: clinicians surrounded by more information than the available processes and tools could comfortably manage. Radiology offered a clear first opening. Imaging volumes were climbing; the number of people available to read the scans was not keeping pace.

A demo, a queue and a hundred refusals

The founders made a demo and showed it to radiologists. The response, as Walach later recalled it, was gratifyingly direct: “I need one of those.” Aidoc's early product looked for urgent findings in medical images and brought potentially time-sensitive cases forward in a radiologist's queue. The ambition was not to remove the physician from the room. It was to make the order of attention smarter.

Investors were less immediately enchanted. Walach says more than 100 venture firms turned Aidoc down. The company did not fit the handy drawers. It was software, but it also faced medical-device regulation. It lived in the cloud, but it had to inhabit old and complicated hospital systems. Category confusion is expensive. The seed round was $3.5 million.

The first installations arrived in 2018 at Sheba and Cedars-Sinai. They made the abstract argument physical. An algorithm could work, but the result still had to appear inside a real sequence of decisions. On the first day of one early go-live, the team received an email about a case that convinced them the work could matter beyond a slide deck. Walach calls it a “one wow moment.” The institution remained a close partner.

In 2019, Walach moved to Miami to be closer to the company's main market. It was geography as operating philosophy. Aidoc's first American hire was dedicated to customer success. Walach and Braginsky, the CTO, took support calls for years. Their product was being educated by its customers, sometimes at inconvenient hours.

The three floors of usefulness

Walach explains a good clinical AI product as a three-layer structure. The first is algorithmic accuracy. Necessary, certainly. Also insufficient. The second is workflow integration: will clinicians encounter and use the result in the rhythm of their actual jobs? The third is the outcome. Did the system change anything worth changing?

The framework explains why Walach now resists describing Aidoc as merely a radiology AI vendor. A hospital cannot sensibly contract with a different company for every finding, then stitch dozens of shallow integrations into one coherent system. Aidoc's strategic answer has been a platform: common infrastructure for deploying, monitoring and governing multiple applications across specialties.

It also explains his skepticism toward model worship. An algorithm can score beautifully and still become a nuisance. It can send too many alerts, arrive without context or ask a busy person to adopt a parallel workflow. The laboratory rewards performance. The workplace is more demanding; it asks whether performance can behave.

Walach's management habits have the same operational cast. He cites Kaizen, the Japanese practice of continuous improvement. He seeks mentors and credits Yuval Bar-Gil with helping him think ahead. Before Walach's first board meeting, the two rehearsed for days. The picture is mildly comic and entirely human: the teenage university student, the air-force AI leader, the billion-dollar-company CEO, still practicing the slides until the joins disappear.

There is a generosity in admitting the rehearsal. Founder mythology prefers the natural performer who strides into a boardroom already fluent in everything. Walach's version is more credible. He treats leadership like software that ships in versions. Openness to correction is not a confession that the previous version failed. It is the maintenance plan.

Half a billion dollars, and another test

By 2026, the scale had changed. Aidoc said 20,000 physicians at nearly 2,000 hospitals relied on its technology. Its first disease-specific product had taken roughly three years to develop, validate and clear. The company was now investing in a clinical foundation model intended to shorten development for broader groups of findings, as well as reporting tools and the underlying operating platform.

In April, Aidoc announced a $150 million Series E led by Goldman Sachs Alternatives, with General Catalyst, SoftBank and NVentures also participating. The round took reported funding above $500 million. Axios reported 31 FDA clearances at the time and a longer corporate ambition: Walach said an initial public offering could come within three to five years.

His nearer aspiration is attached to 2030. By then, he has said, every complex diagnostic decision should be supported by AI. The sentence contains a founder's confidence, but the story behind it supplies the caveat. Support requires accuracy, physician control, governance, integration and trust. The software has to make itself useful repeatedly, not merely announce that the future has arrived.

Walach is fond of hiring for potential over experience. In Aidoc's early days, advisers discouraged him from choosing a business-development leader with no healthcare background. He trusted his reading of the person, made the hire and later turned the decision into a general preference. The anecdote is revealing because it is not an argument against expertise. It is an argument for growth as a form of evidence.

The same standard now applies to the technology he sells. Clinical AI has potential in abundance. Walach's career has been a decade-long attempt to make it earn experience: case by case, integration by integration, call by support call. The algorithm opens the conversation. The patient work of becoming dependable finishes it.