Will Ross arrived in insurance carrying a question about fire. At Stanford, he had been studying how wildfires spread. An insurer had an obvious reason to be interested: a burning hillside can become a collection of claims. A better model seemed like a plausible business. Then Ross and his future co-founder started asking how the people buying those models actually used them. The question grew more troublesome, and considerably more interesting.
There was already a clue to his intentions in an email. William Steenbergen, studying computational mathematical engineering at Stanford, remembers Ross writing that he was tired of ad optimization and wanted work with an impact. The two connected through Steve Blank’s Lean LaunchPad course. Ross’s message stood out among the introductions. Here was someone with technical interests and a reasonably specific objection to spending his working life making advertisements more efficient.
Ross had brought several subjects to the table. At Tufts, he studied philosophy and environmental science. He went on to an MS in climate and atmospheric modeling at Stanford and an MBA at its Graduate School of Business, graduating from the MBA program in 2021. He had worked in IBM’s Watson Group and spent time at Venrock. His background gave him experience with both AI and the companies trying to make a business from it.
The fire research was substantial work in its own right. His paper, Being the Fire, appeared at the NeurIPS 2021 Workshop on Tackling Climate Change with Machine Learning. It used information from 21 western US fire perimeters from 2017, training an agent on one million sub-images. The question was whether reinforcement learning could help predict a fire’s next action. Ross was studying uncertainty at the edge of a burning landscape. Insurance would give him a different edge to investigate.
A very long introduction to the working day
Federato was founded in 2020, while Ross and Steenbergen were still at Stanford. Early conversations with insurers made them look beyond selling a fire model. The modeling question remained useful, but the organizational question kept intruding. A company could buy better information and still struggle to put it into the decisions that mattered. Getting the analysis right was one task. Getting somebody to use it, at the right moment, was another.
The founders pursued the second task with interviews. By July 2021, Federato was describing more than 1,250 hours spent talking with underwriters. That is a formidable amount of conversation for two people building a software company. The figure also gives some scale to a frequently abbreviated part of startup life: learning the customer’s job. Underwriting has its own language, incentives and exceptions. The founders gave themselves time to hear those details.
Ross has described starting with cold LinkedIn outreach and paying some interviewees $100 Amazon gift cards for an hour. There is something pleasantly ordinary about that detail. A company working with advanced mathematics acquired some of its understanding through a digital gift certificate and a willingness to ask questions. The work involved persuading busy people to explain a day they probably had too little time to finish.
In a March 2021 essay, Ross recounted a design colleague interrupting a chief underwriting officer’s discussion of operational issues. The interruption shifted attention toward the people experiencing those issues. Ross then wrote about Jamie, an underwriter handling middle-market business and school referrals, who liked Game of Thrones. The point of introducing her was practical: another pricing tool or distribution initiative could leave her daily difficulties intact. A business problem became easier to investigate when it had a person attached.
“I want to do something I actually care about.”
Ross’s email, recalled by William Steenbergen
Twenty sensible decisions, one awkward portfolio
The problem Ross kept circling was communication. An insurer’s portfolio strategy could sit some distance from its day-to-day underwriting. The organization might know how much risk it wanted, where it wanted it, and what concentrations it wished to avoid. The individual underwriter still needed guidance that reflected what colleagues were doing now. A guideline in a PDF could describe the plan without keeping pace with the portfolio being assembled.
Ross explains the difficulty through decisions that look reasonable on their own. A specialist can accept an account after weighing its merits. Trouble can emerge when other specialists make similar exceptions and nobody has a current picture of the combined result. Insurance accumulates through those choices. The portfolio changes while the queue keeps moving. A decision can be defensible locally and awkward in company.
Federato uses RiskOps as the name for bringing portfolio strategy into underwriting operations. Think of it as a continuing conversation between the overall book and the next account. What the company already holds should influence what it considers next. New decisions should, in turn, update that view. Ross has called the software “a mechanism of communication.” The phrase is modest enough to be revealing.
It also explains why simply adding another score would leave him dissatisfied. A score still needs context: the carrier’s priorities, its existing exposure, the trade-offs it is willing to make. Federato’s task is to carry that context into work. Ross’s interest lies in the moment when information changes a choice. The attraction of the problem is partly mathematical, but its consequences belong to people buying coverage and the people deciding how to offer it.
Company priorities
Next decision
The people who could translate
Learning an industry through interviews also changed who helped run the company. Ross and Steenbergen met Megan Bock Zarnoch during their research into underwriting. She brought decades of insurance experience and became Federato’s COO. Their attempt to understand the customer introduced them to someone who could help make that understanding part of the organization. Customer discovery had produced a leadership connection.
Ross has spoken about the value of insurance practitioners inside sales and customer success, too. An underwriter, broker or actuary can hear a customer’s concern with a familiarity that a technically capable outsider may lack. That understanding matters when an apparently small request carries a history of workflows and business rules. Ross’s experience as an outsider appears to have made him attentive to the need for people who could translate.
Early investors helped with another kind of translation. Pear first met the founders through its Stanford GSB competition and brought them into PearX S20. Its account of the relationship describes work on expressing the idea as operational underwriting software and navigating long enterprise sales cycles. A technical vision had to become something an insurance organization could buy, adopt and use. Understanding that organization took more than knowing how to train a model.
This is a recurring feature of Ross’s career: he spends time at the boundary between a technical idea and its audience. His public biographies credit him with educating more than 1,000 executives through Udacity’s AI for Business Leaders program. At Federato, the audience became more specific. The question was how to make a system useful to an underwriter with an account to assess, a broker awaiting an answer and a portfolio that would survive either way.
Capital, then a wider question
The company’s funding grew alongside its product ambitions. A $40 million Series C led by StepStone Group in November 2024 brought Federato’s announced total to $80 million. In November 2025, it announced a $100 million Series D led by Growth Equity at Goldman Sachs Alternatives. Federato reported that revenue had tripled over the preceding year. Those are company figures, describing the business at the time of the announcement.
Ross’s explanations of AI remained grounded in the limits of a working day. In 2026, he used taxi dispatch as an analogy: a dispatcher trying to coordinate cars with radios, phones and maps has a difficult optimization problem. Software can improve the experience without proving that every trip is the mathematically perfect one. He sees a related opportunity in insurance, where improving the selection and order of work can matter across a portfolio.
By May 2025, he was discussing Orchestrate on NYSE TV, describing an agent-building system that could automate processes such as gathering news about potential insureds. The scope was widening from supporting a decision toward arranging the work around it. In a separate Profiles in Risk conversation, he discussed Orchestrate alongside Submission Triage. The familiar question persisted: what should reach the underwriter, and when?
By September 2026, Federato described a platform extending through policy administration, billing, claims and renewal. Its argument is that these activities need shared context. What happens to a claim can matter to the next underwriting decision; a policy change can matter to billing. Ross’s original concern about disconnected knowledge now reaches across more of the company’s working life. Building across that span gives his team a much larger job than selling a wildfire model.
Keep pedaling
For a description of how he expects that job to get done, Ross reaches for a bicycle. Federato’s September 2026 culture essay says he compares the company to a road bike, with the rider doing the work of climbing and covering distance. It is an unusually serviceable image for a venture-backed company. A funding round may pay for equipment. The gradient still expects an answer from somebody’s legs.
There is a concrete hiring practice alongside the metaphor. Every candidate has a final interview with Ross or Steenbergen, and the hiring committee must agree unanimously. The founders retain a direct role in deciding who joins. After years of asking underwriters to explain their work, they are still spending time in conversations that can shape the company’s future.
Ross’s public calendar also shows the questions moving outward. He was listed for a September 29, 2026, ITC Vegas session about why some insurers get value from AI and others do not. The title suits the route he has taken. His career has given him reasons to care about the models and reasons to ask what an organization actually does with them.
The fire researcher, the executive educator and the founder are recognizable in the same ambition. Ross wants data to have consequences in the working day. Federato began when he and Steenbergen followed that ambition past their first product idea, into a long series of conversations. The software has grown, and so has the company around it. Somewhere in the next queue is another decision, waiting for the rest of the organization to catch up.
