Neomind Learning, led by co-founder and CEO Felipe Hildebrand, wants no part of the “AI tutor” label. Instead, the company is positioning itself around a new term: Personal Learning Intelligence.
“‘AI tutor’ has become a catch-all for any AI-powered services (most of which is a chatbot) when applied in educational contexts,” Hildebrand says. “However, a true tutor is far more than a chatbot on guardrails. A great tutor carries knowledge mastery and delivers it very effectively by building trust and coaching the student through a productive struggle, so needed for true learning. That is what we built at the very core of Neomind.”
Two Systems, One Architecture
Personal Learning Intelligence rests on two systems. The first is the Learning Genome, which Hildebrand calls “a map of how each child reasons, learns, and engages” — one that can “craft every single lesson individually to each student” and “sharpens with every interaction.” The second is the Instructional Mentor, “a teacher designed from the first day to feel human: it reads engagement, builds rapport, and responds with the warmth and intuition of a great educator.”
“We start from the learner needs and build outward from there,” Hildebrand says. “That’s the difference between a chatbot with guardrails and Personal Learning Intelligence.”
He draws a sharp line against conventional adaptive-learning tools, which he says typically adjust only pace. “The content stays the same; students just move through it faster or slower,” he says. “Even when students are stuck on the same problem, they may be struggling with very different concepts. One child’s learning gap opened three lessons ago in trigonometry; another’s opened three years ago in fractions.” Because the Learning Genome persists across sessions, he says, “the system isn’t meeting a stranger every session. It’s picking up a relationship, the way a great tutor would.”
Welcoming the Skepticism
Hildebrand doesn’t push back on scrutiny of AI companies serving children. “That skepticism is fair, and honestly, I think the industry needs more of it,” he says. “Too much of what gets built with AI in education optimizes for looking finished at the surface rather than building the underlying robust learning fundamentals. In this space, the bar has to be higher, because we are talking about children, not conversion rates.”
The company spent roughly a year and a half building and testing before any families gained access, and the caution continues now. “The second half of this year is still a learning phase for us: sharpening what we say, listening to what actually helps families, and building toward independent studies with research partners, so our claims stand on evidence, not enthusiasm,” he says. “I would rather be slow and honest than fast and hollow.”
The Access Argument
Chief Learning Officer Dr. Tracy Ocasio, who spent nearly four decades in public education, frames the mission around access. “The single biggest constraint I saw everywhere was time,” she says. “A great teacher in a classroom of 30 students simply cannot give every child the individual attention research tells us actually moves the needle. Private tutoring can, but it has historically been expensive and out of reach for most families, including many of the Hispanic, Black and English-language-learner communities I’ve spent my career serving.”
She’s careful to frame the technology as a complement, not a substitute. “Nothing replaces a great teacher. That’s not the ambition here,” she says. “If we can extend even a fraction of what a hundred-dollar-an-hour tutor offers to families who could never access that, at a price point that’s actually reachable, that’s a meaningful step toward leveling the playing field.”
Why Voice, and Why Compliance Matters
Co-founder and CTO Pedro Forli says the gap AI can fill isn’t good teaching but one-on-one time: “AI can bridge that gap and provide personalized learning experiences that we just don’t have enough humans (teachers) to deliver.” Cost, he adds, is central too: “Good tutoring is expensive, but AI can supplement education at a lower cost.”
Neomind also bet on voice over text. “Most tools in this space are built as text-based chatbots, which have real limitations in how engaging and instructionally effective they can be,” Forli says. “We made an early bet on voice as the core interface, because conversation is closer to how real tutoring happens.” Chatbots, he argues, simply can’t run a class the way a tutor does: “Chatbots cannot ‘host’ a class, as they don’t know when you finished a particular topic, and they are extremely limited in what they can gather from how you are reacting to the instructions.”
On data, Forli says the company builds “around two federal frameworks, COPPA and FERPA, as the floor, not the ceiling. For students under 13, that means parental authorization before any AI interaction, clear data retention and deletion policies, and strict limits on what happens to student data.” That standard shaped the architecture directly: “The biggest architectural decision this drove was keeping all data processing inside our own infrastructure rather than routing it through third-party AI vendors, since student data can’t leave our environment under those compliance standards.” Student interactions aren’t used to train models, he adds: “Data is only used for support and quality purposes, it’s anonymized, and personally identifiable information is redacted before it’s stored.” As more states introduce their own rules around AI and minors, he says, the company’s approach is to “build to the strictest applicable standard rather than adjust state by state.”
Spencer Hulse is the Editorial Director at Grit Daily. He is responsible for overseeing other editors and writers, day-to-day operations, and covering breaking news.
