JACKIE
Project Collaborators
Project Overview
The Jacket Project (Jackie) was conceived as a digital assistant to empower neurodivergent girls and women by providing clear, adaptive, and trustworthy daily guidance—particularly around clothing choices and routines.
The intention was to build an app specifically tailored to neurodivergent needs, addressing challenges such as overwhelming interfaces, vague instructions, or excessive setup. The design concept combined weather-driven clothing recommendations, adaptive support modes (get dressed, head out, activities), and safe chat functionality powered by Azure Cognitive Search and AI speech.
The prototype envisioned using machine learning models (YOLO4, EfficientNetB1)to suggest clothing, while adapting over time to user preferences. The planned reminders could be delivered via icons, text, or voice, and the interface was intended to be clean, inclusive, and accessible—prioritizing privacy and local data storage.
Envisioned Key Features
- Weather-based Clothing Recommendations - Jackie would analyze live weather data to suggest practical, appropriate outfits.
- Three Support Modes - “get dressed,” “head out,” and “activities,” each guiding the user through context-specific routines.
- Safe and Trustworthy Chat - RAG ensured answers were drawn from preselected online resources, reducing the risk of misinformation.
- Multi-Modal Support - users could interact via icons, text, or voice depending on their preferences and comfort level.
- Routine Support - notifications or even printed QR codes could serve as visual cues to help users start routines at the right time.
- Adaptive Learning - Jackie was designed to learn user preferences over time, tailoring clothing suggestions and reminders.
- Neurodivergent-Centered Design - clean user interface, simple setup, and explicit instructions, avoiding the overload of mainstream apps.
Planned Technologies
- Azure Cognitive Search - used for retrieval-augmented generation (RAG) chat, ensuring that Jackie’s responses came from curated, trusted sources.
- Azure AI Speech - to allow voice-based interaction, making routines easier to start without needing to navigate menus.
- YOLO4 & EfficientNetB1 - computer vision models for recognizing clothing items and predicting weather-appropriate outfits.
- Bildstöd Icons - Swedish pictorial support icons to provide explicit, easy-to-understand instructions, particularly valuable for neurodivergent users.
- Frontend + Backend Integration - a lightweight frontend paired with cloud services to manage chat, recommendations, and user settings.
- Local Data Storage - user preferences and personal data were intended to be stored locally for privacy and safety.
