Image
Image

Sparky: an AI-first study buddy for children

Design

AI Core

EdTech

Voice Agents

Backend

Engineering

AI hardware for children, homework support, voice games, child-safe agent

Sparky is an AI device concept for children aged 6–11.

The device helps with homework, supports voice games, keeps a child-friendly dialogue, and works through school tasks at the child’s pace. The character lives on a small always-on display and can switch between study, play, and daily companion modes.

We worked on the market research, product concept, device architecture, technical solution, landing page, digital twin, CAD direction, and psychological framework for child-agent interaction.

Problem.

Children need help between school, homework, and parents’ availability

Homework support often depends on whether an adult has time, patience, and enough context to help.

For children, the difficult moment is usually small and immediate: a missed concept, a confusing exercise, a reading task, a math mistake, or a need to ask the same question again without feeling judged.

Sparky addresses that daily gap through a dedicated device. The child can talk to it, upload homework, play voice games, and get help without turning every task into screen time on a phone or tablet.

Product concept.

Study buddy, voice game partner, morning alarm

Sparky was shaped as a physical AI companion with several modes:

Sparky was shaped as a physical AI companion with several modes:

  • homework support,

  • voice games,

  • conversation,

  • learning prompts,

  • daily routines,

  • morning alarm.

The device is designed to feel present in the child’s room or study area. It has a soft physical form, a small screen, animated character states, and direct voice interaction.

The site explains it simply: “Study buddy. Party guest. Morning alarm.”

Technical challenge.

Continuous character animation with near-zero operating cost

The hardest technical constraint was cost.

The device needs to show seamless character animation and speak at any moment. At the same time, the base experience should not force users into an expensive monthly subscription just to make the product usable.

That constraint shaped the architecture. We needed a setup where the device could keep animated states running, respond through speech, and support interaction without relying on costly generation for every moment on screen.

This affected agent logic, animation strategy, speech behavior, local-state design, and backend usage.