Post

Our Hackathon Project in Impact Hack 2025: CozySteps

Posted
Preview Image
By Chencheng Liang
1 min read
Our Hackathon Project in Impact Hack 2025: CozySteps

CozySteps — Our LLM-powered activity recommender for families with autistic children 🧩

Built during Impact Hack 2025 by AI Sweden

🧠 Background & Motivation

We joined Impact Hack 2025, a hackathon organized by AI Sweden that encourages teams to use AI-driven innovation to tackle real societal challenges.

Our chosen challenge:
help families with autistic children discover suitable activities for their kids.

Many autistic children are highly sensitive to sensory input and have narrow interests. Parents often struggle to find appropriate activities, which can make family interactions stressful.
With CozySteps, we aimed to reduce decision fatigue for parents and make daily life more enjoyable for everyone.

🤖 What We Built

We created a mobile app prototype called CozySteps,
with an activity recommendation engine powered by Large Language Models (LLMs).
It analyzes each child’s profile and activity metadata to provide personalized suggestions.

Key features:

  • 🧒 Child Profile Management
    Parents can record the child’s age, interests, and sensory sensitivities.

  • 🎯 LLM-powered Activity Recommendations
    A curated activity database combined with an LLM inference layer suggests 3 tailored activities at a time.

  • 💬 Parent Feedback Loop
    After completing an activity, parents can give simple feedback. The LLM interprets feedback to refine future recommendations.

  • 📅 Schedule & Reminders
    Parents can add activities to a calendar and receive reminders to build predictable daily routines.

The UI is calm and icon-based to minimize sensory overload and support neurodiverse users.

📸 A Glimpse of the App

CozySteps app interface

🛠 Tech Stack

  • Frontend: React + styled-components
  • AI logic: LLM-based reasoning + rule-based activity metadata
  • Data: local JSON storage
  • Prototyping: Figma

We chose a lightweight front-end + local data + LLM API inference architecture to stay focused on user experience and AI logic during the hackathon timeframe.

Participating in Impact Hack 2025 was an inspiring experience —
we’re happy we could build something warm and helpful for families with autistic children 🌈

Events
This post is licensed under CC BY 4.0 by the author.