Research

Graduate Research Intern

Child & Family AI Learning Companion

About us

We're second-time founders and former Google and Synthesia leaders (Search, Voice, and Assistant). We're assembling a core team of engineers, artists, and educators to build the interface between AI and childhood.

The problem we're solving

Kids between 6 and 10 spend hours every day on screens, but most digital content is passive (TV, YouTube) or repetitive (apps, games). Parents want safe, engaging, and educationally positive experiences. Kids crave play, parents crave development and safety, and the industry hasn't bridged that gap.

Our product

Pebble is building the first generative learning world. If you've read the Diamond Age, you know what that means. If you haven't, this video will bring you up to speed.

The first version of Pebble is an exploration game. Children explore a procedurally generated world alongside an AI companion they interact with through voice. Conversations are open-ended, not scripted. The companion isn't a sidekick, it's the core relationship that drives engagement, learning, and emotional investment.

Founders

Guillaume (CPTO): First product hire at Synthesia, where he scaled from $3 to $100M ARR and helped define the category of AI video. Spent 7 years at Google as a PM on Search. Dad of two (ages 7 and 9). LinkedIn.

Lio (CEO): Second-time entrepreneur, founder of Neoplants (raised $25M+ from top-tier US investors). Ex-Google PMM, part of the original Google Assistant EMEA launch team. 15 years across tech, storytelling, and consumer launches. LinkedIn.

About you

You're a graduate student who finds children genuinely fascinating: the way they reason, the questions they fixate on, the leap from "why is the sky blue" to two hours of self-directed astronomy. You suspect the next decade of childhood will be shaped by what AI companions get right or wrong, and you want a hand in figuring out which one it is.

You're rigorous about evidence but allergic to research that sits in a drawer. You like products. You like kids. You want to spend a summer watching real families use a real thing and shaping what we build next.

The role

We have a core hypothesis: Pebble can rapidly identify a spark of interest in a child — dinosaurs, unicorns, rockets — and, over time, transform that spark into productive inquiry: reading, math, scientific reasoning, sustained attention. Your job is to help us prove, disprove, or sharpen that hypothesis with US families.

You'll work directly with the founders and the product team on a research practice that feeds straight into the roadmap. Concretely, you will:

• Run family interviews and home visits. Watch 6–12-year-olds use Pebble in their living rooms. Talk to parents about what they hoped for, what surprised them, what worried them.
• Evaluate learning activities. Assess our reading and math content against learning-science principles, US curricular expectations (e.g. Common Core), and against what actually engages children in the wild.
• Audit AI companion interactions. Review transcripts and recordings for safety, tone, and learning quality. Flag patterns, good and bad, that should shape our prompting, guardrails, and pedagogy.
• Help design progression experiments. Translate the interest-to-inquiry hypothesis into testable studies: what does "productive inquiry" look like in week 2 vs. week 8, and how do we measure it?
• Synthesise insights for the team. Turn observations into opinionated memos that the team can act on next sprint.

Requirements

You have

• Current enrolment in a US-based graduate programme in HCI, learning sciences, child development, psychology, education, cognitive science, or a closely related field.
• Hands-on experience running qualitative research with children and/or families — interviews, observation, think-alouds, diary studies.
• A working grasp of how US children aged 6–12 develop literacy, numeracy, and curiosity, and the ability to read learning-science literature critically.
• Strong written communication. You can turn 20 hours of footage into a one-page memo with a clear recommendation.

You're comfortable with

• Working in an early-stage startup. Methodology trade-offs are constant; perfect studies are the enemy of useful ones. You ship insights at the cadence of a product team, not a thesis committee.
• Holding two stances at once: child safeguarding is non-negotiable, and we're building something genuinely new that won't fit existing rubrics. You can navigate both.
• Forming opinions. We don't want a neutral observer — we want someone who'll tell us when an activity is pedagogically thin or when the companion is being sycophantic.

Nice to have

• Prior work on edtech, children's media, or AI tutoring systems.
• Experience with mixed-methods evaluation (combining qualitative observation with engagement/learning metrics).
• Familiarity with US safeguarding and privacy frameworks for children's digital products (COPPA, FERPA where relevant).
• Ability to recruit families through your own networks (schools, clinics, community groups).

Genuinely optional

• A favourite picture book you'd defend in a fist fight.

Work conditions

• Paid internship, 3–6 months, full-time preferred (part-time considered for the right candidate).
• US-based. Remote within the US, with a strong preference for candidates who can run home visits and family observations in person in their region. Travel budget for fieldwork.
• Must be authorised to work in the US.
• Direct mentorship from the founders; your work lands in the product, not a slide deck.