Yair Adato and
the price of a picture

At eight, Yair Adato fell in love with mathematics and decided he wanted a doctorate. Children have announced less practical plans over breakfast. What makes this one interesting is where it eventually took him: into the problem of teaching a computer to understand a picture, and then into the problem of deciding who should be paid when a computer makes one.

Adato is the co-founder and CEO of Bria AI, established in 2020. Its business joins image generation to licensed training material and a system for attributing value to the people behind that material. In a field understandably preoccupied with what appears on the screen, he keeps returning to the arrangements underneath it.

There is a personal continuity here. Mathematics offered structure. Computer vision offered a way to investigate images through structure. Bria asks whether a creative industry can grow around generated images with an economic structure that includes their human contributors. The questions have become bigger; the taste for working out how the pieces fit has remained.

The boy who liked the numbers

His academic home was Ben-Gurion University of the Negev. His research addressed motion estimation and the reconstruction of reflective shapes, with work involving Harvard University and an internship at Google. In 2011 he received a Google Europe PhD Fellowship in Computer Vision. The childhood ambition had acquired a rather specific vocabulary.

Consider a shiny object. Its surface shows the surrounding world in distorted form. Understanding that surface from an image means separating information about the object from information about what it reflects. Adato’s research with Ohad Ben-Shahar examined how motion and shape could be recovered together. A reflection, which seems to offer everything at once, can make the underlying object remarkably difficult to understand.

That research does not make his later business inevitable. It does establish that his involvement with images predates the recent enthusiasm for generating them. He had spent years considering what an image reveals, what it conceals, and what a machine might extract from it. A convincing visual result was already a problem to investigate rather than simply admire.

An education among the shelves

At Trax Retail, the scene became less abstract. The company applied image recognition to retail, where understanding a shelf means identifying products and turning what a camera sees into information a business can use. His 2015 conference presentation concerned fine-grained detection in retail auditing: availability, assortment, visibility, pricing, and promotions.

There is little room for romantic vagueness in a shelf audit. A shop needs useful information about what is present and what is missing. The work connects a technical capability to a repeated commercial task. An algorithm has to earn its place in the routine.

Adato progressed through technical leadership roles, including VP of research and development and CTO, as Trax grew from a small startup into a much larger international business. He calls the experience “my real-world university.” It is an apt description for a researcher learning what happens when infrastructure, customers, and accountability enter the same room.

Some relationships survived the change of company. About a third of Bria’s team had come from Trax by August 2025, including chief technology officer Michael, or Misha, Feinstein. That is a tangible connection between the two businesses: colleagues bringing their experience of making computer vision useful into the next venture.

“It was my real-world university.”

Yair Adato on Trax

A picture with people behind it

Bria began in 2020, before image generators became a familiar public pastime. Adato’s early argument was that visual communication demanded too much specialist skill, time, and money. Someone might know the picture they wanted yet lack the means to produce it. Generation could shorten the distance between an idea and an image.

The company also chose to build around licensed material. That choice puts a relationship with rights holders near the beginning of development. Before an image becomes training data, there is a permission question. After the resulting technology earns money, there is a participation question.

Bria’s attribution engine is intended to connect generated outputs to contributions within the training data and support compensation accordingly. Adato is a named inventor on a US patent granted in July 2024 concerning attribution of generated visual content to training examples. Alongside him are Ran Achituv, Eyal Gutflaish, and Dvir Yerushalmi.

The patent describes analysis of properties in generated content and training examples, linking an output with a group of examples, and recording their contribution. The distinction matters. A generated picture need not be a pasted collection of existing photographs for the work used during training to have economic importance.

Getty Images became a strategic partner in October 2022. Other data relationships included Envato and Alamy. For a founder arguing that creative material should enter AI through agreement, those relationships are part of the product’s substance. They supply both training material and a route for rights holders to participate.

Samsung Next joined Bria’s $24 million Series A in 2024. In March 2025, Bria announced a $40 million Series B led by Red Dot Capital, bringing the announced total raised at that point to $65 million. These are company financing milestones. Their relevance to Adato’s story is the backing they provided for an approach that requires more than assembling an appealing image interface.

His preferred commercial territory has been inside other people’s workflows. Bria offers developers ways to integrate generation and editing capabilities into their products. That suits an executive who describes his interest as bridging technology and business. A person using a creative tool need not know every component behind it for that component to matter.

The right to change one thing

By November 2025, another concern had taken a prominent place in his argument: revision. An image can be attractive and still fail a brief. A creative team may want different lighting while retaining the composition. The ability to request a picture is useful; the ability to make a particular correction has a different practical value.

Bria’s FIBO model approaches that problem through structured descriptions in JSON, a format software can read and modify. A short instruction can be expanded into explicit visual properties. The system returns an image and its structured description, giving the user a record to inspect and refine.

The company presents this as a route toward more predictable generation. Lighting, camera angle, palette, and composition become properties to adjust. The intended benefit is easy to recognize without knowing the format: a revision should follow the instruction closely enough for someone to build a working process around it.

“Teams can’t build workflows around randomness, or scale what they can’t control,” Adato said when FIBO launched. The sentence sounds like the Trax education resurfacing. A model’s usefulness depends on whether its behavior can fit the job being asked of it.

The name FIBO comes from the Fibonacci sequence. For someone who recalls discovering mathematics at eight, the choice is a pleasing return to familiar territory. Numbers have made their way back into the picture, this time in the name of a tool intended for creative work.

Getty’s current description of its own AI services names Bria as its technology partner and describes a custom model trained on licensed content and proprietary data. It gives Adato’s infrastructure argument a concrete setting: the technology can reach customers through a visual business they already use.

The house has other management

Outside the company, Adato has described a household with his wife, twin daughters, and pets. A 2022 speaker biography gave the cat and dog the authority to run the house. By February 2026, he was living in New Jersey and describing family time as a way to stay grounded. His public LinkedIn profile places him in Tenafly.

These small details do useful work in a career otherwise crowded with models, licensing arrangements, and technical vocabulary. They locate its protagonist in ordinary life. Even a founder concerned with control can apparently be overruled at home by an animal with no interest in the roadmap.

He has also written about the field in public. His 2022 essays considered the coming wave of generative AI and changing assumptions about images and video. Another tackled visual search in digital asset libraries. Even the subjects of his writing suggest someone interested in the daily inconvenience between having an image and being able to use it.

His public activity has continued beyond still images. In June 2026, Bria announced a new version of its video background removal model. An April appearance on The Media Copilot returned to licensed data, trust, and creator economics. The scope of the tools changes; the recurring questions remain recognizable.

Adato has also offered other founders a practical piece of advice: distinguish a technology’s future promise from its present capability. A pilot can attract enthusiasm while failing to deliver continuing business value. It is a restrained observation from someone whose company depends on a field with considerable expectations attached.

His ambitions include carrying attribution and control into more visual media and opening commercial opportunities for rights holders. Those remain ambitions to be judged through products and working relationships. The career so far gives them a particular shape: research that asks how an image works, operating experience that asks what it is useful for, and a company that asks how its contributors can share in the proceeds.

The eight-year-old wanted to work with numbers. The adult has chosen a business where a picture’s price must account for people who never appear in it. That is a sufficiently demanding equation to keep him occupied.