Neo.Tax co-founder and CTO Firas Abuzaid has watched as so many industries try to force AI into places where it just doesn’t fit: a hammer-in-search-of-a-nail situation.
But it’s become clear to him that tax is a space where machine learning is the solution. “The Big 4 firms effectively substitute data with reputation. They dispatch expensive teams of accountants to interview your employees and handle all the paperwork for you,” he says. “But you can build a solution that takes in a ton of data from a business, understands the business's needs and context, and uses that to generate an output that is human-readable; that the IRS can actually digest and say, ‘Yes, this makes sense.’ And it's going to be an AI solution.”
From Stanford to Startup World
During undergrad, Firas studied computer science at Stanford. He went on to get both a Master’s and Ph.D. in machine learning and systems in Palo Alto. He was fascinated with database research, but around the time of his graduate research, “big data and machine learning became kind of synonymous, because you couldn't actually do any sort of valuable ML without big data.” So, Firas started diving into answering foundational questions: “How do you train models efficiently? How do you run inference on data efficiently? How do you make sure that you're not sacrificing quality when you do all that stuff?”
During his Ph.D., he took an internship at Microsoft Research that was focused on computer networking. Hired to optimize traffic on wide-area networks, Firas started to think that a career at Microsoft Research would be in his future. But around that time, Neo.Tax founder Ibrahim reached out to ask if he’d like to work on a startup idea to automate the R&D tax credit.
Firas had been pitched many a cofounder position by that point, but was impressed with Ibrahim and his idea. “I told him, ‘Thanks, but no thanks,’ but we just kept in touch as friends.” But then, COVID hit. And, by that point, Firas was no longer as optimistic about a career as a researcher at a big industrial lab. He was frustrated, sheltered in place, stressed. “Like a lot of folks, the pandemic just kind of forced me to reassess my whole life,” he says. “I took a step back, and I was like, ‘I don't feel very happy. I don't feel very fulfilled. I could try something new.’”
Right then, Ibrahim reached back out and showed Firas the progress they’d made. The startup, then named cpa.ai, had an MVP by then and a few customers signed up. Ibrahim asked if Firas would be open to putting in 10 hours per week. “I started working with him for 10 hours a week, and then 10 hours became 20, and then 20 became 40, and then, before you knew it, we were applying to YC, and we were raising a Seed Round,” Firas remembers. “Once the momentum began to build, it just took off.
Tax as a Big Data Problem
Coming from the world of research and academia rather than accounting gives Firas a distinctive outlook that’s allowed Neo.Tax to take off. Right away, he understood that the problem of business tax meant sorting and optimizing massive amounts of data according to complex rule sets. But it also required understanding a business’s needs and circumstances at a deep level to produce the most accurate results. “This is going to sound surprising to people, but taxes, especially for big businesses, is a big data problem,” he says. After spending a decade studying big data, he knew ML could help Neo.Tax tackle the big data of taxes.
So, what does Firas mean by Tax as a Big Data Problem?
To file correctly, you have to understand all the transactions, all the expenses, all the work that's being done, all the employees, the org structure, and more. “You have to reconcile data across so many different sources, and piece it together to build this cohesive picture of what the business is trying to do,” he explains. “If you don't do that, then you're going to have a massive blind spot, which leads to a big inaccuracy in what you're filing with the IRS.”
After years of work, it’s been exciting for Firas to see Neo.Tax start to really soar. He still believes the product has a ways to go; he knows there are more problems to solve and the system could be even more efficient. “But we've found a real pain point that doesn't go away, except with AI, right? I think that's really the key,” he says. “We found a problem where AI is necessary to win, and the AI has to be really high quality. We've done the hard work of making our AI viable.”
The Art of CS
I ask Firas if he ever expected to be here, in a managerial role, using his Computer Science degrees to tackle business taxes. “No,” he says through a grin. Mainly, he explains, he never thought he’d be a manager, because he loves to “dive deep” to figure things out. “But I do think having the ability to dive deep makes me a better manager.” But also, there’s a part of him that is not surprised he’s ended up in an industry he never expected.
See, getting here, to the role of CTO at Neo.Tax, makes sense for a student who loved to try a bit of everything in school. He was always a great pupil, but he could never settle on just one place to devote his focus. He was drawn to math and science, but also was obsessed with arts, humanities, and international relations. “I was trying to find different subjects that would whet my appetite in these respective ways,” he says. “I was kind of split between left and right brain.”
It wasn’t until his first CS class at Stanford that Firas found his calling. For the first assignment, his professor assigned something basic to the 700 enrolled students and said to them, “The program we’re going to ask you to write is extremely simple, in a highly constrained environment. When you read the assignment description, you will think, ‘this seems pretty easy to do.’ But it’s going to be harder than you think. And, unless you cheat, no two submissions will be identical. Everyone is going to have a unique solution.”
For Firas, coding became the arena he’d been subconsciously searching for: an expertise that was both art and science, and that scratched both sides of his brain. (Many others in the field agree with him, by the way.) “I realized that you can creatively express yourself in all these different ways when you're writing code, because of how you organize your thoughts,” Firas says. “If I asked you and 99 other people to create a painting of a mountain, you would get 100 different mountains. And it's kind of the same thing with code.”
For most people, there’s a Pacific Ocean-sized gap between the romance of painting and the drudgery of business taxes. But for Firas, taxes have proven to be an especially thorny big data problem to solve. So, the Ph.D. from Stanford started to paint a mountain; now, a half-decade in, his AI solution is helping Neo.Tax find its way to the top.
Catch up on the latest news and updates
Subscribe To Our Newsletter
Insights on R&D tax credits and AI innovation delivered to your inbox every month.