MindFlayer

Inspiration

Many children and elderly people struggle with modern digital interfaces. Typing, complex menus, and small screens can make technology frustrating instead of helpful. At the same time, AI systems are becoming more powerful, but they are often not designed with accessibility, safety, or age-appropriate interaction in mind.

I wanted to build a voice-first AI companion that feels friendly, simple, and safe — something that children can use to learn and ask questions, and elderly users can rely on for daily assistance and companionship without needing advanced technical skills.

This idea led to MindFlayer, a voice-based AI buddy focused on accessibility and human-centered design.

What it does

MindFlayer is a voice-powered AI assistant designed specifically for:

Children (learning, curiosity, simple explanations)
Elderly users (daily reminders, clear answers, friendly conversation)

The system works entirely through voice interaction:

Users speak naturally using a microphone
The AI responds in short, calm, and easy-to-understand sentences
The assistant adapts its tone and behavior based on the selected user mode (Child or Elder)

Safety and clarity are core priorities. The assistant avoids harmful, medical, or unsafe advice and encourages users to consult parents, caregivers, or professionals when appropriate.

How I built it

The project was built as a simple full-stack web application:

Frontend

HTML & CSS for a clean and accessible interface
JavaScript for handling voice input and output
Browser-based Speech Recognition for capturing user voice
Text-to-Speech for spoken AI responses

Backend

Node.js with Express to handle requests
A lightweight API that processes user input and mode selection
Prompt logic designed to be compatible with Google Gemini (AI.dev)

AI Integration

The AI behavior is controlled through carefully designed prompts
Separate prompt logic is used for Child Mode and Elder Mode
Responses prioritize simplicity, politeness, and safety

Due to temporary accessibility issues with AI.dev in my local environment, I validated and demonstrated the same prompts using an alternative AI testing platform. The prompt logic and structure are fully transferable and designed to work with Google Gemini models.

Challenges I faced

Platform accessibility: Some AI lab tools were unavailable in my local setup, so I had to adapt by using a compatible alternative while keeping the design aligned with Google AI principles.
Safety design: Designing AI responses that are helpful but not unsafe required careful prompt wording.
Voice interaction: Making voice input and output feel natural while keeping the system simple was a key challenge.
Time constraints: Building a complete, demo-ready project within a short hackathon timeframe required focusing on core features and clarity over complexity.

What I learned

Voice-first design can significantly improve accessibility
Prompt engineering plays a critical role in AI behavior and safety
Clear explanations matter as much as technical implementation
Honest prototyping and transparent documentation are essential in hackathons

What’s next

Future improvements for MindFlayer include:

Full deployment with Google Gemini APIs
Multilingual voice support
Smarter reminder and scheduling features
Optional parental or caregiver controls
Integration with smart devices for hands-free assistance

MindFlayer demonstrates how AI can be designed to be inclusive, safe, and human-centered, especially for users who are often overlooked by traditional interfaces.