When Lax Poojary was at Google, he spent years building consumer products meant to make information easier to access – travel tools, short-form video apps, shopping experiences, you name it. But it wasn’t until he started watching how children used screens that a different problem began to bother him.
Kids were overflowing with curiosity, but the tools meant to teach them rarely kept pace, offering answers without interaction or room to explore.
A child might ask how a city could be built on Mars – a question that mixes science, engineering, economics, and imagination – and the internet’s answer was usually a mainstream YouTube video or a block of text. Informative, maybe. Interactive? Rarely.
The gap felt more than obvious. The technology existed to make learning dynamic, yet children were still stuck consuming static content.
And it was that frustration that became the starting point for Sparkli, a Zurich-based startup building what it describes as a multimodal, AI-native learning engine for children ages 5 to 12.
The goal is not to digitize textbooks or nestle AI into existing curricula, but to give kids a system that responds to how they actually think – by talking back, showing outcomes, and letting them experiment.
Poojary founded Sparkli alongside two longtime collaborators from Google, Mynseok Kang, now chief product officer, and Lucie Marchand, the company’s CTO.
All three had worked on experimental products inside Google and its Area 120 incubator, where they learned firsthand how hard it is to build tools people return to, as well as how quickly users disengage when products feel rigid or outdated.
Education technology, they felt, was what really had that problem at scale.
“Children learn by exploring, making choices, asking questions, and discovering what inspires them,” Poojary says. “Sparkli turns screen time into a place where curiosity grows rather than fades.”
Turning questions into quests
Sparkli’s answer is an AI system built around real-time exploration. Instead of following a fixed lesson plan, children start with a question or idea – whether that be building a city in space, running a food cart, or designing a sustainable neighborhood – and the platform generates an interactive learning path on the fly.
Voice, visuals, simulations, and decision-making tools are layered together so kids aren’t just watching or reading, but actively steering what happens next.
Ask Sparkli to build a city on Mars, and it doesn’t return a paragraph. It walks the child through trade-offs, including gravity, energy sources, oxygen production, and housing design. Each choice leads somewhere different, forcing the child to reason, revise, and explain their thinking.
This design reflects what the founders call the “agency gap” in children’s digital learning. Screens are everywhere, but most of the time kids are passengers, not drivers.
Under the hood is an AI-driven interest and knowledge graph that tracks what a child explores, where they struggle, and how their thinking evolves. That data shapes future experiences, gradually increasing complexity while staying aligned with the child’s interests. The emphasis is on skills development – problem-solving, creativity, systems thinking – rather than memorization.
The approach has so far attracted early interest. Sparkli is currently running a pilot with one of the world’s largest private school groups, including roughly 100,000 students. In classrooms, teachers are using the platform less as a lesson replacement and more as a launchpad.
When kids take the lead
In one pilot session, eight-year-old students used Sparkli to simulate running a small food cart. The exercise quickly moved beyond math into pricing strategy, customer experience, and sustainability, all prompted by the kids’ own decisions. The role of the teacher shifted from instructor to guide.
Building this kind of system for children, however, comes with constraints that don’t apply to general-purpose AI tools. Open-ended chatbots can confuse young users or surface inappropriate content, which is why Sparkli’s founders made an early decision to tightly constrain the system, embedding guardrails that shape not just what content is allowed, but how it’s presented.
To do that, they’ve assembled a team that spans AI research, education science, game design, and motion graphics, including contributors affiliated with ETH Zurich. The challenge is balancing generative AI’s flexibility with the structure required for safe, age-appropriate learning – especially under regulatory frameworks like COPPA in the U.S. and GDPR in Europe.
That focus on trust has resonated with investors, with Sparkli recently raising $5 million in a pre-seed round led by Founderful and securing the runway needed to expand pilots and deepen the product.
Looking ahead, the team sees Sparkli as a foundation to a system that could grow with a child over years, adapting as interests shift from dinosaurs to design, from space to social systems.
