Big Tech companies and upcoming startups want to use generative AI to build software and hardware for kids. A lot of those experiences are limited to text or voice, and kids might not find that captivating. Three former Google employees want to get over that hurdle with their generative AI-powered interactive app, Sparkli.
Sparkli was founded last year by Lax Poojary, Lucie Marchand, and Myn Kang. As parents, Poojary and Kang were not able to satisfy their children’s curiosity or give engaging answers to their questions. Poojary explained that while he would use tools like ChatGPT to explain concepts to his six-year-old, the result was still just a wall of text. He noted that what kids truly want is an interactive experience, which became the core idea behind founding Sparkli.
Prior to launching Sparkli, Poojary and Kang co-founded a travel aggregator called Touring Bird and a video-focused social commerce app, Shoploop, at Google’s internal startup incubator, Area 120. Poojary later worked at Google and YouTube on shopping. Marchand, the CTO of Sparkli, was also a co-founder of Shoploop and later worked at Google.
Poojary described the evolution of learning tools, saying that fifty years ago a child asking about Mars might have been shown a picture, and ten years ago a video. With Sparkli, the goal is to let kids interact and experience what Mars is like.
The startup believes education systems often fall behind in teaching modern concepts. Sparkli aims to teach kids about topics like skills design, financial literacy, and entrepreneurship by creating AI-powered learning “expeditions.” The app lets users explore predefined topics or ask their own questions to create a learning path. It highlights one new topic daily. Kids can listen to a generated voice or read text. Chapters within a topic include a mix of audio, video, images, quizzes, and games. The app also creates choose-as-you-go adventures that remove the pressure of getting questions right or wrong.
Poojary mentioned that the startup uses generative AI to create all of its media assets on the fly. The company can create a learning experience within two minutes of a user asking a question and is working to reduce this time further.
While AI assistants can help children learn, Sparkli notes their focus is not primarily on education. To make its product effective, the company’s first two hires were a PhD holder in educational science and AI and a teacher. This was a conscious decision to ensure content serves children better, keeping pedagogical principles in mind.
Safety is a key concern around kids using AI. Sparkli said that while topics like sexual content are completely banned, when a child asks about sensitive topics like self-harm, the app tries to teach emotional intelligence and encourages talking to parents.
The company is piloting its app with an institute that has a network of schools with over 100,000 students. Its current target audience is children aged 5 to 12, and it tested its product in over 20 schools last year.
Sparkli has also built a teacher module that allows teachers to track progress and assign homework. Inspired by Duolingo, the app is designed to be engaging so kids want to return frequently. It uses streaks, rewards, and quest cards based on a child’s initial avatar to encourage regular learning.
The startup has seen positive responses from school pilots. Teachers use Sparkli to create expeditions for class discussions or to assign exploratory homework to gauge understanding.
While Sparkli plans to primarily work with schools globally for the next few months, it aims to open consumer access and let parents download the app by mid-2026.
The company has raised $5 million in pre-seed funding led by Swiss venture firm Founderful. This is Founderful’s first pure-play edtech investment. The firm’s founding partner, Lukas Weder, said the team’s technical skill and market opportunity prompted the investment. As a father, Weder sees Sparkli as a product that can engage kids in immersive learning about relevant topics, away from just video games.
This post was first published on January 22, 2026.