Three People Who Met Through an Accelerator Quietly Built the UK's First Computer Vision Unicorn.

This is the story of Alex Dalyac, Razvan Ranca, and Adrien Cohen, three founders with completely different backgrounds who came together to build Tractable. This AI platform has helped millions of people recover faster from accidents and turned a $55 million raise into a $1 billion company.

Alex Dalyac: The Economist Who Fell in Love With Deep Learning

Building an AI company was never something Alex Dalyac aspired to do growing up. Dalyac grew up in France, academically inclined, wanting to see how things worked from the bottom up, studied Econometrics at LSE – an area that requires Mathematics, statistics, and Economics. In Econometrics, he learned how to identify and spot quantifiably predictable patterns in complex, messy data sets.

From LSE, Dalyac moved on to complete a conversion Master's in Computer Science at Imperial College London. It was during this time that he was first exposed to the concept of deep learning. While still an emerging area in academia at this point, Dalyac immediately knew he wasn't seeing an abstract research concept; he was witnessing a tool with potential to affect real-world problems on a level unprecedented by anything before it.

He went on to apply deep learning to its first industrial purpose at Imperial, leading the Computing department. This alone conveys the pace at which Dalyac had transitioned from discovering deep learning to utilizing its capabilities in the real world. It didn't matter to him to study the technology for purely academic purposes; he had an ambition to know what deep learning could do when applied to an actual problem.

Upon graduation from Imperial, Dalyac went on to briefly work as a quant for a hedge fund and then as one of the first employees at Lazada, the rapidly growing e-commerce venture expanding through the whole of Southeast Asia. In each of these roles, he gained even more insight and conviction that technology, when applied correctly, could compress decades worth of development within just a matter of years; yet, both roles still didn't feel right.

He wanted to build. He wanted to apply the knowledge of deep learning to a real issue he could affect, and he needed the right partner to help him build that company.

It was at that point that he partnered with Entrepreneur First.

Razvan Ranca: The Cambridge Machine Learning Mind Who Wrote Poker Bots

Razvan Ranca grew up in Romania, went to the UK to go to university, and was doing research work that a good scientist does their entire life to get. He studied Artificial Intelligence and Computer Science at the University of Edinburgh, arguably the strongest AI program in Europe, and then got an MPhil in Cambridge. He managed to convince one of the UK's premier machine learning professors to advise him.

At Cambridge, he wasn't just getting a degree; he was writing algorithms. He was writing poker bots. He was doing work at the bleeding edge of what ML could do and gaining an intuition about what ML could reliably do and what it couldn't, which would serve him well a couple of years later at Tractable, helping him determine which real-world problems it was possible to solve with deep learning, and which were not.

He gained a distinction for his MPhil and then chose, somewhat surprisingly to everyone around him, to forgo further academic research, or a job in Big Tech, and join Entrepreneur First to attempt to start a company.

"I wanted to apply my skill-set to a novel set of problems that are currently intractable," he later explained. He did not choose his words lightly.

At Entrepreneur First, he met Alex Dalyac. Two individuals who, at that time, both independently concluded that deep learning was the most important technology of the day, and building companies applying it to problems was a more interesting prospect than studying the field further, and that there was a moment in time to go for it and do it. They recognized one another immediately.

The Entrepreneur First Meeting That Started Everything

Entrepreneur First is a talent investing platform where you bring super talented people, often just fresh out of graduate programmes, and give them the time, the resources, and the community to find a co-founder and build a company. When Alex and Razvan met each other in 2014 at the Entrepreneur First programme, it was still early days, and the culture was in its early stages of forming what it meant to make a high-stakes bet on yourself when you didn't have the established credibility that most investors would typically look for.

They had it. They chose to ignore the clear career paths that they offered to them. That orientation towards the problem was the beginning of their partnership.

They explored a variety of problem spaces. Pipe corrosion. Earthquake imagery. Medical imaging, in particular X-rays. They were trying to find a deep learning computer vision application that was extremely technologically viable, scalable from a commercial perspective, and provided real human benefit.

Then they considered damaged vehicles.

The insurance claim for a damaged car in 2014 was an almost completely manual task. A car was damaged. A claim was made. A loss adjuster was dispatched to inspect the car in person. The loss adjuster would write up a report, and then the cost estimate would be negotiated. This took days or weeks, frustrated everyone, and cost insurance companies more money in order to carry out. And importantly, this was a visual problem. The entire assessment of damage rested on someone's ability to see the damage and estimate the severity, the cause of the damage, and how much it would cost to repair.

Deep learning computer vision was, in 2014, becoming really powerful for visual tasks. It can be trained on images of damaged vehicles from millions of examples of what the damage looked like and how much it cost to repair. The technology was ready. The market was enormous. The human benefit was obvious, where people who were in accidents would get their cars repaired faster.

Alex and Razvan looked at each other and said the same thing. This was the problem.

"We realized that this was the perfect market," said Adrien Cohen, who would later join them in the following year as the third co-founder.

Adrien Cohen: The Goldman Sachs Banker Who Had Already Built One Unicorn

Adrien Cohen wasn't around for the start of Tractable when Entrepreneur First had ended. He came after the seed round; he was the experienced operational and commercial man whom Alex and Razvan freely admitted to needing.

He looked so different from the founders. Educated at a French grande école, former investment banker at Goldman Sachs, and then what we would now call his defining prior career move, joining Lazada following a recruitment approach from Rocket Internet. This was the e-commerce platform setting itself up across SE Asia in an attempt to mirror the Amazon business in Vietnam, Indonesia, Thailand, and the Philippines – all rapidly changing, challenging, and fundamentally not like the Western internet companies for which the startup playbooks had already been written.

It was at Lazada in Vietnam where he and Alex first worked together and developed a mutual respect that had sustained through the years that passed before he arrived at Tractable.

Lazada had been bought by Alibaba in 2016 for $3 billion in one of the biggest tech buy-outs to occur in SE Asia to that date, and Adrien Cohen had been one of those to build a company worth three billion dollars before he was 40.

When Alex called him about Tractable, Cohen knew precisely what was being asked. The science and technology were there; what was required was somebody who knew how to walk into a major global insurance company and win them over, and close a commercial relationship – because he'd already done that before. Somebody Alex would later describe as "knew how to talk to executives".

"He was incredibly valuable in enabling those key relationships to be formed," Alex would say of Cohen. "He was the one providing accelerated supervised learning."

It is this last phrase, lifted directly from the same machine learning vernacular the company itself was founded on, that should be taken note of. Alex Dalyac thinks in models, and, in describing what Adrien Cohen brought to him and Razvan, it was the language of the technology they were creating; he could provide supervised learning and could accelerate his and Razvan's learning in areas that they hadn't yet learned to model themselves.

First Milestone: Winning the Trust of the World's Largest Insurers

Launched commercially in 2014, Tractable spent its early years patiently and methodically building the evidence base that a large insurance company would demand before handing its claims process to a system that it had never seen before.

Its first customers were referrals from Cohen's network and direct, dogged perseverance from a team that was trying to persuade extremely conservative, risk-averse companies to change an essential part of how they operated. Insurance companies did not rush: proof was demanded, not promises.

The proof, when it came, was impressive. Tractable's AI could assess the damage to a vehicle in photographs in minutes, not days. The AI was consistent, unlike a human assessor whose opinions depended on experience, fatigue, and local knowledge; the AI could handle volume far beyond that of a human team, and it constantly improved, becoming more accurate with each dataset.

The first large UK insurer, then Aviva, then a whole collection of international giants (Tokio Marine, Sompo, Ageas, and, in America, GEICO) joined the platform. More partnerships mean more data, which in turn means the AI model is more accurate, which in turn brings more partnerships. The flywheel had started.

By 2020, the company was on CB Insights' list of 100 global leaders in AI. It won Best Technology at the British Insurance Awards, and its system was assessing over $2 billion of vehicle repairs and purchasing per year.

The Funding Story and the Path to Unicorn Status

The journey for Tractable had been one of measured development and conservative growth relative to its industry. It took seven rounds of fundraising and $55M to reach unicorn status.

In June 2021, the company raised a Series D funding round of $60M, which valued it at $1bn; this made it a unicorn and the first computer vision unicorn from the UK to achieve the milestone, and the first company to reach unicorn status through the Entrepreneur First accelerator.

In July 2023, Tractable announced a further $65M Series E round, which brought the company's total funding up to $184.9M. This round was led by SoftBank Vision Fund 2, and the investment was significant for its network in the insurtech, auto, and property sectors worldwide; with such networks, it opened commercial avenues that money itself would not.

Alex stated, "We've turned the $55M into a $1B company," referring to the 2021 unicorn valuation, and said that this had been achieved through discipline, a narrow focus on delivering the correct product, and the capital efficiency of the system they built.

The Leadership Transition and Where Tractable Stands in 2026

In July of 2024, Alex Dalyac stepped down as CEO of Tractable after roughly 10 years at the helm of the company, turning a business from two founders in an accelerator into a global unicorn with offices in London, New York, and Tokyo, that had worked with over 20 of the world's biggest insurance companies.

Razvan Ranca continues as CTO, in charge of the direction of the company. Its technology now looks towards including generation AI and large language models in the backbone of the Tractable platform, which aims to expand beyond vehicle damage assessment into property damage and wider disaster/accident response to become an infrastructure provider for the entire insurance industry when physical damage needs to be managed.

Adrien Cohen stepped down from his role as an active co-founder. His commercial foundations, the initial work and relationships with global insurance companies in the very early, vital years of the company, remain embedded in the foundations of the partnerships that form the driving force of Tractable.

Tractable continues to process millions of claims a year, speeding up the time it takes for an insurance claim on a natural-disaster hit house to be assessed (from weeks to hours) or for a damaged car on an accident scene to be repaired (speeding up the process to get a customer back driving).

What Their Story Teaches Every Founder

Three founders, three very different career backgrounds, and one conviction about a technology and a problem. Tractable is a case study in complementarity taken to its most professional conclusion. Alex had the academic underpinning and deep learning know-how; Razvan, the engineering rigor, ML research, and insight as to what the technology could confidently achieve; Adrien, the commercial nous, executive contacts, and operational expertise. Alex and Razvan could not have sold a powerful but unproven technology into the financial services industry without Adrien. At the same time, Adrien would have struggled to have brought anything to market without them. A measured and incremental scale is the second lesson. In a conservative industry like insurance, credibility stems from delivered performance, and not hype. Tractable's founders did not try to scale the business ahead of what the technology could actually do; they focused on getting it to work well, allowing that result to convince the next stakeholder to listen.

Final Thought: Once Intractable Problems

Their name said everything about what they aimed to achieve. Tractable. Able to be dealt with. Manageable.

In 2014, the idea that an AI system could take an image of a wrecked car and spit out a stable, commercially sound, objective assessment of damage was not tractable to industry practice. That was research. That was an interesting line of inquiry. Something that would likely be achievable in due course.

Alex Dalyac, Razvan Ranca, and Adrien Cohen believed it to be achievable now and gathered the evidence to demonstrate it. Millions of people's claims were settled more quickly, and cars were back on the road faster than they would have been without three people deciding, at an accelerator programme in London, that this was tractable in 2014. That's what it looks like when the right people identify the right problem at the right time.