He left Georgia Tech in the middle of a pandemic, built a video chat company, sold it, and now runs the thing his four-year-old parents' generation would have called an impossible sentence: an AI co-teacher for the American classroom.
Ask Sohan Choudhury how Flint got good and he will tell you, without a lot of curated humility, that early on he felt like he was selling snake oil.
That is the word he used. Snake oil. In interviews. Out loud. The problem, in 2023, was that most AI-in-schools pitches were pretty demos to district administrators who liked being seen as forward-thinking. Choudhury and his co-founder Jinseo Park had built something. They were selling into an excitement about a technology, not into a measured effect on ninth graders learning algebra. There is a version of this story where a founder gets rich, ships slides, and never notices. Choudhury got on planes instead. He kept getting on them. He sat in the back of classrooms with a notepad and watched teachers use the product, watched kids poke at it, watched the specific ways it fell over. Flint's roadmap started to sound less like a chatbot and more like a curriculum. Then the revenue doubled.
By November 2025 Flint had closed a $15 million Series A co-led by Basis Set Ventures and Patron, with USC Viterbi, AME Cloud Ventures, Afore Capital, Y Combinator, and Matt Pittinsky - the co-founder of Blackboard, which is the education software company you likely used in college and definitely resented - writing angel checks. Four hundred thousand students and teachers were using the thing. A hundred and fifty thousand AI-powered learning activities had been generated. The investors who spent 2024 telling Choudhury that education doesn't scale had, mostly, stopped saying it.
Flint is AI-native, not retrofitted. It's built for schools, with educators, to personalize learning at scale.Sohan Choudhury, to AlleyWatch, November 2025
Flint is an AI platform for K-12 schools that lets teachers personalize instruction without turning it into a second job.
Teachers upload their materials - a worksheet, a rubric, a novel chapter - and Flint spins up interactive, subject-specific activities that meet each student where they are. Struggling students get scaffolded through the problem. Advanced students get pushed. Everyone gets real-time feedback that a single adult in a room of thirty could not otherwise deliver at the speed of a class period. The bet is that AI-in-schools is not really a chatbot bolted onto homework. It is a system that sits underneath the whole day.
Interactive practice built from a teacher's own materials, differentiated per student in real time.
Teachers see what students actually understood, not just what they marked on a form.
Custom rubrics, AI-driven assessments, and feedback loops calibrated to the teacher's standards.
Education is the great mobilizer.Choudhury, on why he and Park are building for schools
Choudhury calls this Flint's "greatest Achilles' heel." He says it out loud. He said it in the EdTech Chronicle Q&A. He says it in podcast interviews. He does not soften it. Two engineers, ex-Georgia Tech, no classroom hours, trying to build the operating layer for American K-12 instruction. The obvious critique is that this is exactly the sort of hubris that has produced fifteen years of edtech that teachers politely ignore.
Choudhury's response has been unglamorous. He and Park started visiting schools in person - not for a demo, for a school day. They watched. They rewrote features that looked elegant in Figma and broke in a real fifth-grade classroom. The pattern held: whenever Flint hit resistance, the response was more field time, not more marketing. Teachers, it turns out, notice.
Co-founder of Blackboard.
"The biggest challenge, arguably, was that the majority of investors have preconceived notions about the potential scalability of an education product."
"Many tools claim to personalize learning. Few are built from the ground up in close collaboration with schools and teachers."
"To teach every student in their class based on their needs, not just to the median level."
Sohanchoudhury.com is not a normal founder site. It has a list of things he has done, a link to his Twitter, and an offer to "15 minute chat sike" - a joke about the standard Calendly link every VC-adjacent 26-year-old CEO puts at the bottom of a personal site. His favorite location, according to a company Q&A, is "The Flint Office." His Twitter handle is @hungrysohan. His parents immigrated to the United States in the late 1990s and he cites that, unprompted, when reporters ask why he cares about education specifically. All of this reads a little contrived until you spend time on the actual product, at which point it starts to read like consistency.
Public, active, occasionally about food.
Publicly noted the internal process was a lot.
Google and MIT were paying customers. He was 21.
Our greatest Achilles' heel is that neither of us has ever taught.Choudhury, in EdTech Chronicle
Ask Choudhury what Flint should look like in five years and he tends to use the words "pioneer" and "new product category." He wants Flint to be the platform every school reaches for when the question is not "how do we ban ChatGPT" but "how do we teach every kid to their own level." That is a large sentence. It is also a specific one. The old model of education software - which is roughly Blackboard, which is roughly the LMS - was infrastructure for delivering the same lesson to everyone. The bet Flint is making, and it is not the only company making it but it is one of the more coherent ones, is that AI collapses the cost of one-to-one instruction to something a public school district can pay for. If Choudhury is right, the shape of a normal school day changes. If he is wrong, he still built a serious business and learned something.
He talks like someone who thinks he is right. He also talks like someone who has been wrong before, at a smaller company, and remembers what it felt like. That combination is more interesting than either of the two on their own.