Inspiration
Early childhood is full of moments that feel unforgettable while they are happening, but quickly become scattered across photos, messages, calendars, notes and memory.
Traditional baby books usually ask parents to reconstruct those moments later. Baby Kangoo started from a different question:
What if families could capture small moments as they happen, and AI could help turn those fragments into a meaningful record over time?
We wanted to create something that combines the practical side of following a baby's first years with the emotional value of preserving the family's story.
Just as importantly, we wanted AI to have clear boundaries.
AI may improve the storytelling, but it may never invent the story.
What it does
Baby Kangoo is an AI-powered early childhood companion designed around a baby's evolving journey.
Families can record daily memories, moods, milestones, first foods, vaccinations, pediatric appointments and meaningful “firsts”. The experience adapts to the baby's age and combines these private family records with curated age-based information about development, nutrition and vaccination schedules.
Gemini powers two core experiences.
Kangoo Insights
Kangoo Insights analyzes the family's own recorded memories over a selected period and organizes them into grounded summaries, recurring themes, highlights and useful reminders for future pediatric conversations.
Every AI observation is tied back to real family records. Baby Kangoo does not diagnose, evaluate developmental delay or provide medical advice.
Parents can revisit previously generated analyses without spending another AI call and download them as PDF snapshots.
Kangoo Stories
Kangoo Stories turns memories into a growing family memory book.
Rather than asking Gemini to create emotional stories from sparse data, Baby Kangoo first guides parents through contextual memory questions. Their answers become new factual source material.
Gemini can then connect those memories into a warm narrative while remaining grounded in what the family actually recorded.
The result is a book built chapter by chapter, where AI helps families remember more before it writes more.
How we built it
Baby Kangoo was built as a mobile-first web application using Next.js and TypeScript, developed with Google Antigravity.
Supabase provides authentication, persistent family data and Row Level Security, isolating every family's records. Demo sessions are also isolated, allowing judges and new users to explore Baby Kangoo safely without creating a full account first.
Gemini, through the official Google GenAI SDK, runs exclusively server-side.
The AI architecture follows a grounded pipeline:
authenticated user → baby ownership verification → Supabase records → sanitized context → Gemini → structured response → evidence validation
Internal temporary evidence references allow Baby Kangoo to verify that AI observations correspond to records actually supplied to the model.
Curated reference information such as age-based guidance, nutrition content and country-specific vaccination schedules is stored separately from personal family data. This keeps factual reference content deterministic while Gemini focuses on interpreting the family's own memories.
The application is deployed through Vercel, with GitHub used for version control and isolated development branches.
Challenges we ran into
One of the biggest challenges was deciding what AI should not do.
A model can easily turn a short memory into a beautiful story by inventing weather, emotions, dialogue or scenery. For a family memory book, that would make the writing richer but the memory less truthful.
Our first Stories prototype made this trade-off very visible. Strict grounding produced writing that was accurate but sometimes too descriptive.
That led to one of the most important design decisions in Baby Kangoo: instead of relaxing the grounding rules, we introduced guided memory interviews.
The solution was not to let AI invent more details. It was to help the family remember more details first.
We also had to solve persistent anonymous sessions, cross-user isolation, dynamic baby age calculations, AI response caching, country-specific health content and preventing automated tests from polluting production data.
Accomplishments that we're proud of
Baby Kangoo evolved from an idea into a working deployed product during the hackathon.
We built:
persistent real and demo family profiles diary, mood, milestone and timeline systems country-aware vaccination guidance age-based development and nutrition experiences real Gemini-powered Kangoo Insights with evidence grounding persisted AI analyses with period-based caching downloadable Insights PDFs grounded Kangoo Stories guided memory interviews persistent chapter-based family memory books RLS-protected family data a production deployment accessible on the web
We are especially proud that the AI features are not decorative additions. They change what families can do with the memories they record.
What we learned
The most important lesson was that AI becomes more useful when its role is narrower and clearer.
Baby Kangoo does not need Gemini to generate vaccination schedules or developmental reference content. Those areas benefit from curated, deterministic information.
Gemini becomes valuable where human memories are fragmented and personal: finding patterns, organizing moments, asking better questions and turning confirmed details into meaningful narratives.
We also learned that good AI UX is not simply about generating an answer. Persistence, transparency, loading states, evidence, user control and the ability to revisit previous outputs matter just as much as the model response itself.
What's next for Baby Kangoo
The next step is expanding Baby Kangoo to a small group of real families and pediatric professionals for structured feedback.
We plan to add growth tracking with visual growth charts, expand country-specific developmental and health reference content, and continue building tools that help parents arrive at pediatric appointments better informed and better prepared.
Over time, Baby Kangoo could become a bridge between everyday family observations and pediatric care: helping parents organize growth, milestones, questions and relevant history in a way that is easier to understand and share with their child's doctor.
We also plan to expand guided memory-book chapters and explore multi-caregiver family accounts.
Longer term, Baby Kangoo can become a private, evolving archive of childhood, combining the emotional story of growing up with a structured record of development and care.
The goal is not for AI to replace pediatric judgment or write a child's childhood. It is to help families preserve what they lived, understand what they observed, and bring better context into the conversations that support their child's development.
Built With
- antigravity
- api
- gemini
- genai
- next.js
- postgresql
- react
- sdk
- supabase
- typescript
- vercel
Log in or sign up for Devpost to join the conversation.