Ajay Krishna Amudan has never seen much use for an answer that is almost right.
As a college student at IIT Madras, he competed in programming contests, where the scoring is unforgiving. At the ICPC world finals, which his three-person team reached in Marrakech in 2015, a solution that fails even one of the judges’ hidden tests earns no points. He had loved math since childhood for the same reason. “There’s no almost or kind of correct,” he says.
A year ahead of him at IIT Madras, Ram Krishnamurthy was teaching a computer to play Mario. DeepMind had recently shown that an AI could teach itself simple Atari games. Ram’s research took on longer games, which can’t be won by reacting one move at a time. “If you want Mario to decide to jump, you can’t just go up to the last thing and jump,” he explains. The work was published, but Ram returned from a summer research internship at USC knowing that he “didn’t want to stop at just writing papers.” What interested him was what research could change outside the lab.
The two now run Hyphenate (formerly Maximor) which builds AI agents to do the work of a finance department: sending invoices, reconciling cash, closing the books at the end of each month, and more. Their goal is for Hyphenate to become the company finance teams run on: one bigger than SAP.
Today, its agents are taking on finance work at customers including Groupon and Synthesia. Once one workflow proves itself, customers tend to hand over another. The average customer now runs six of the modules in its platform.
Ram and Ajay believe the path past SAP is to change how work happens. SAP and legacy systems of record like it track what a finance team has done. Hyphenate does the work itself, with evidence attached to every step, so it ends up with a live, verified picture of how money moves through a business. Ram and Ajay see this as the foundation for moving up into strategy: helping CFOs answer harder questions, like whether pricing should change or where the next dollar of R&D should go.
In the meantime, Hyphenate has outgrown three offices in the same WeWork building near Bryant Park, going from eight seats to fifteen to thirty. Last week it signed for a fourth.
Hyphenate’s name on the Nasdaq tower, with its founders, Ram and Ajay, below.
Early influences
Ram grew up in Chennai as an only child, which he believes shaped him "in good ways and bad." His mother is a doctor. His father is the CFO of a public company in India. Academics mattered, but his parents weren’t pushy. They expected excellence and left the definition of it to him. At school, his classmates elected him head boy, the student chosen to lead the rest.
Every year during audit season, his father disappeared into work. Ram was drawn to the business behind the numbers.
"I always had this fascination of ‘how do you build a company from scratch?’" he says.
Ajay also grew up in Chennai. As a kid, he followed Roger Federer and Cristiano Ronaldo down to the most obscure statistics. He looked up to them for how they played as much as for how often they won. The Ronaldo of 2005, he points out, was “a beautiful player to watch,” the kind who would do a trick just to see what was possible. Federer shared Ronaldo’s love of the game. “The word I’d use is grace,” Ajay says. “He took victories and defeats the same way.”
Sports shaped how he thinks about competition. To him, an opponent isn’t an enemy but a reason to push harder. “You want to do better than them, and you take pride in that,” he explains.
His first math competition was the Math Olympiad, when he was about nine. It was supposed to be a proof-based exam, which Ajay didn't know. He worked the problems out on scrap paper and turned in a single sheet of answers. By the rules, he should have been disqualified. He placed in the top ten in southern India.
He found mathematics “partly beautiful and partly truthful,” and planned to study it until his last two years of high school, when he was introduced to algorithms. The earliest one he remembers is binary search, which finds a number in a sorted list by cutting the list in half again and again. In a list of a million numbers, it needs only about twenty guesses. The solution was simple and elegant, and it was also useful.
That pairing pushed him from pure math toward its applications. At IIT, he studied computer science and spent his time on two things: theoretical computer science, writing his thesis on a problem called matroid intersection, and competitive programming.
Ram with a photo of himself and his dad, a public-company CFO in India.
Free radicals
While still at IIT, Ram co-founded FastNext, a startup that applied AI to recruiting, and served as its CTO. It was acquired in his final year. He’s clear-eyed about why the outcome wasn’t bigger. The team had built a strong product but underinvested in selling it. He resolved not to make that mistake again.
He graduated with two offers. One was a place in Stanford’s management science and engineering program. The second came from Raja Krishnaswamy, a Microsoft exec who had come to the IITs to recruit a handful of graduates directly to the U.S. Ram told him he wanted risk and room to build, not a career ladder. “Use me however you want,” he said, “but I want to be a free radical.” Raja offered him a place on a new team doing Palantir-style transformation work for Microsoft’s largest customers, Walmart and Coca-Cola among them. Ram deferred Stanford and, six months in, he gave up his spot.
Raja hired one other graduate from India that year: Ajay. He was put on Azure’s DNS team, which handles billions of queries a day, and he calls the experience a "trial by fire." By the end of 2018, he had done well enough to make a demand. He told the VP of his group that he wanted time without a manager to find a problem he cared about. "If you’re willing to give me six months, then I'll stay. If not, I'm going to leave Microsoft." The VP said yes.
The problem he landed on was billing. Microsoft's cloud was charging customers in increasingly complex ways, and every bill had to be both right and traceable to the usage behind it. The work had what he loved about math: a correct answer, and a way to prove it. It also touched his other obsession. He was reading quarterly reports, day trading and considering a career in quant finance, and he admired what Stripe and Plaid were building.
Over the next two years, the system Ajay built grew from covering $10 to $20M of revenue to roughly $150B. About 80 teams adopted it, up to Microsoft’s CTO. It was, he says, the best period of his time at Microsoft.
Nights and weekends
Ajay doesn't speak Hindi. He credits that, in part, for his friendship with Ram.
They knew each other at IIT. Ram was a year ahead, and they had met a few times. When Raja picked them both for Microsoft, neither got a U.S. work visa in time, so the company sent them to Vancouver for a year. Ajay arrived in November, and the cold, gloomy weather was a shock. Making friends took time too. Many of the other recruits from India spoke Hindi together. Ajay speaks English and Tamil. "That's what first brought me close to Ram," he says. "Both me and Ram just spoke in English."
They discovered they liked the same things. They went bowling, played squash and cricket, and went to arcades, movies, and dinners. They first worked together when Ram moved into data work. He built a data product, and Ajay’s DNS team migrated onto it.
When their visas came through, both moved to Seattle, and their friendship deepened. Through 2020, they met every weekend. On Friday nights and Saturdays, they took business ideas apart from first principles. They also talked about ethics and about what makes people act the way they do. Ajay had worked under some difficult managers, and Ram had found a mentor in Raja. Both came away knowing what kind of leaders they wanted to be.
“You can be extremely ambitious while still being a good person,” Ram says.
Along the way, they noticed they approach problems from opposite directions. Borrowing from computer science, Ram calls Ajay “a depth-first person,” someone who goes to the bottom of one question until it’s solved. He’s “more of a breadth-first person” who likes to survey the whole problem first. Each has a blind spot, and Ram thinks the partnership works because each covers the other’s. “I look at it breadth-first, he looks at it depth-first, and we get to the optimal answer faster.”
Ram and Ajay at Intuit Connect in October 2024, the year they started Hyphenate.
In August 2021, Ram asked Ajay to start a company with him. Ram had moved back to India, and Ajay was home in Chennai on a visit. Ram hadn’t talked about starting a company with anyone else. His pitch was simple: “Trust me, let’s do this together.” Ajay already felt his time at Microsoft was coming to an end. “He just needed a catalyst,” Ram says.
Navigating the idea maze
They spent the next year and a half talking, and in early 2023 began trying ideas in earnest. “We went through a very structured idea maze,” Ram says. “Where do we have an understanding? Where can we talk to people, build a prototype?”
Their first idea was in marketing. Image-generating AI was taking off, and they set out to turn everyday shoppers into influencers. They built products and found real customers. But talking to marketers, Ajay says, “I could just feel that this wasn’t for me.” He also decided marketing was the wrong place for AI. The results were too subjective and too slow to show up.
They moved closer to the domain they knew best: finance. They tried automating tax advisory work for accounting firms. The AI models available then couldn’t hold enough information in their context at once. “We were probably a little ahead of our time,” Ajay says. They turned from accounting firms to finance teams at large companies. There, they found, “it’s not just the most technical problems. Everything is a problem.” Teams were overworked, turnover was high, and the work piled up at the start and end of every month.
Buyers in finance are among the most risk-averse in business. Ram contrasts it with fields where a solution that’s 99% right is good enough. A marketing email can be a little off. A revenue number can’t. Finance also has its own logic, shaped by auditors, regulators, and the particular way each business runs, and few AI teams had spent time inside it. Ram expected it to be one of the last fields AI would reach, and only once models could clear a much higher bar.
At the same time, what makes finance hard also makes success provable. In marketing, Ram and Ajay found, you could never quite show the AI was working. In finance, the books either reconcile or they don’t.
In 2024, they started Hyphenate (then called Maximor) with the conviction that AI, properly built, could take on finance work and would only get better. They also set out to cover the full function. While most finance software is a set of separate tools for expenses, revenue, and the monthly close, Hyphenate is a single platform. That’s Ram's breadth.
Underneath it is Ajay's depth. Rather than sending engineers to interview each customer and create custom agents, Hyphenate's system learns from a team’s spreadsheets, emails, and past work. AI prepares the work and humans review it, the way a senior accountant reviews a junior's. The platform works where teams already live, and nothing has to be ripped out or replaced.
Climbing the next hill
Hyphenate recently closed a $45M Series A, where Foundation followed on its seed check.
For both founders, the round is a means to something bigger and more ambitious. They’re playing a long game, and they measure progress over years, not quarters. Ajay, for his part, is never fully satisfied, even after a successful fundraise. He’s already looking for a higher hill.
Ram also measures the round by what it means for the people who took a chance on Hyphenate. One of them is a founding engineer who joined on the understanding that the founders would find a way to bring him to the U.S. Last year they got him an O-1 visa. Ajay wants at least the first 100 employees to make life-changing money.
Their early customers took a chance too. Ram can list what they did with the time Hyphenate gave back. One took his wife to Croatia. Another took a longer paternity leave, and a third took up sailing. Ram remembers how buried his father was during each month’s close, and he finds stories like these especially motivating.
Before the seed round, the founders used their own money to make early hires, and there were moments when they weren’t sure they’d make it. In a lot of founder relationships, Ram says, that’s when trust starts to break down. “Here there’s just an implicit trust.” He doesn’t hedge about it. “I would only start this company with Ajay.”