-
-
Architecture Overview
Inspiration
The inspiration behind AIM – Adaptive Intelligent Moral Story Teller comes from a very personal place—my 3-year-old daughter.
Like many children today, she began watching videos on platforms like YouTube, and over time it became a daily routine. While such platforms can help children learn quickly through engaging content, I often noticed something interesting: the recommendation systems kept suggesting very similar types of videos, even as she grew older and her curiosity evolved.
This raised an important thought for me—shouldn’t children’s learning content grow with them? Instead of repeatedly watching the same genre of videos, I felt there should be progressive content that adapts to a child’s age, interests, and understanding.
At the same time, I strongly believe that children absorb values and behaviors very quickly. That made me think: what if moral values and good habits could be taught through fun, interactive stories instead of passive videos?
Another motivation was to reduce excessive screen time. Instead of children constantly watching visual content, I wanted to encourage listening, imagination, and interaction, similar to traditional storytelling but enhanced with modern AI.
With the capabilities of the Amazon Nova family of models, I envisioned an AI-powered storyteller that could create adaptive, humorous, and moral-based stories. The system would not only personalize stories as a child grows but also gently use analytics to help improve listening skills, comprehension, and moral understanding—without creating any sense of competition or pressure.
What it does
AIM is an AI-powered storytelling application that delivers interactive, adaptive, and value-based story experiences for children while encouraging listening, comprehension, and moral learning.
Interactive AI Storytelling The application uses the conversational capabilities of Amazon Nova Sonic to conduct engaging storytelling sessions. The AI narrates humorous, moral-based stories and interacts with children through voice or chat, creating a dynamic storytelling environment rather than passive content consumption.
Personalized Story Experiences The system collects basic child profile information such as age, interests, and preferences during registration. Based on these inputs, the application generates age-appropriate moral stories tailored to the child’s interests and developmental stage.
Interactive Learning Through Q&A At the end of each story session, the AI asks fun and thoughtful questions related to the story. This encourages children to listen carefully, think about the story’s meaning, and reflect on the moral values presented.
Conversation Memory & Adaptive Learning Each storytelling session is summarized using Amazon Nova Lite. The summarized interaction is then stored for future use, enabling the application to remember previous sessions and adapt future stories as the child grows and develops new interests.
Cloud Storage & Data Persistence The application stores full chat interactions in Amazon S3 for reference and future analysis. Key structured data such as story summaries, parent and child registration details, child interests, and session metadata are stored in Amazon DynamoDB for fast and scalable access.
Learning Analytics & Insight Dashboard Children’s responses to story-based questions are captured and analyzed to generate learning insights. These insights are presented through a dashboard designed not to create competition or performance pressure, but to help parents understand how their child is improving in: 🎧Listening skills 📖Story comprehension 🌱Moral understanding 🎭Engagement with storytelling
How we built it
- Web Application Layer
- Parent Registration & Login: Interface for parent users to register and authenticate
*Child Registration *: Interface for child and related info registration
Login & Auth Microservice
Responsibilities:
- User registration (parent/child)
- User authentication
- Token generation (JWT)
- Session management
Database:PostgreSQL/MySQL
Data Stored:
- User credentials
- Authentication tokens
Prompt Engine Microservice
Responsibilities:
- Process user prompts
- Conversation flow control
- Response generation coordination
Integrations:
- AWS Nova Lite: AI model for prompt processing
- AWS RDS: Prompt metadata and context
- AWS S3: Chat history and conversation logs
Features:
- Prompt optimization
- Context retrieval
- History management
Analytics Engine Microservice
Responsibilities:
- Usage analytics
- Performance metrics
- User behavior tracking
- Reporting and insights
Database: PostgreSQL
Data Stored:
- Usage statistics
- Performance metrics
- User activity logs
- System health data
- AI Nova Sonic Layer
Responsibilities:
- Advanced AI processing
- Audio/voice processing
- Complex AI operations
Integration: AWS Nova Sonic (AI model)
Use Cases:
- Voice recognition
- Audio generation
- Advanced AI features
Challenges we ran into
Time, a steep AI learning curve, model availability challenges, and building a stable working demo were the key hurdles during development.
🔄 Transition from Java to Python Coming from a strong Java development background, moving to Python for AI development required adapting to a different programming style, libraries, and ecosystem. This transition involved learning new frameworks and development patterns commonly used in AI-based applications.
🧠 First AI Project – Steep Learning Curve This project marked our first hands-on experience building a generative AI application. Understanding concepts such as prompt design, conversational flows, adaptive storytelling logic, and integrating AI models required significant experimentation and learning.
⚙️ Limited Access to the Reasoning Model During development, access to Amazon Nova Lite was not yet enabled in our account. This caused delays while implementing the conversation summarization and reasoning components. Temporary workarounds were used during development, with plans to fully integrate the model once access becomes available.
💻 Inconsistent Behavior of the Voice Model While testing the interactive storytelling sessions using Amazon Nova Sonic, the model sometimes behaved differently across development environments and laptops. This required additional troubleshooting around configuration, dependencies, and runtime setup.
Accomplishments that we're proud of
Reimagining Everyday Activities with AI One of our biggest accomplishments was the thought process behind this idea—recognizing that AI is becoming part of everyday life and exploring how it can be used meaningfully to improve simple activities like storytelling and learning for children.
Stepping Out of the Comfort Zone This project pushed us to step outside our comfort zone, learn new technologies in the AI space, and quickly adapt to evolving tools and frameworks. More importantly, we were able to turn this learning into a working prototype within a short time.
From Idea to a Real Impact Vision Building this POC has motivated us to take the idea further and potentially convert it into a real application that can help children learn moral values, listening skills, and curiosity through fun and adaptive storytelling.
What we learned
This project taught us not only technical skills but also the value of curiosity, adaptability, and execution:
🤖 AI and LLM Integration Through this project, we gained hands-on experience working with generative AI models, integrating LLM capabilities into an application, and designing conversational flows for interactive storytelling.
🔄 Adapting to New Technologies The project reinforced the importance of staying open to new technologies. It required quickly learning and adapting to emerging AI tools, frameworks, and development approaches.
🛠 Turning Ideas into a Working Prototype We learned how to move from a concept to a tangible implementation by building a functional prototype that demonstrates the core idea and its potential impact.
⏱ Time Management and Focus Working within limited time encouraged us to prioritize features, manage tasks efficiently, and focus on building a meaningful proof of concept.
What's next for AIM - Story Teller (Adaptive Intelligent Moral Story Teller)
The vision for AIM is to build an AI storyteller that learns with every interaction and grows alongside the child.
There is a lot more to explore and build. This proof of concept has shown that the idea is achievable, but several enhancements are needed to transform AIM into a fully functional and impactful application.
Prompt Engine Refinement The next step is to refine the prompt engine to improve storytelling quality, moral delivery, and conversational flow. Integration with Amazon Nova Lite will be prioritized to enable better summarization and reasoning for conversation memory.
Advanced Analytics Enhancing the analytics engine to provide deeper insights into children’s listening patterns, comprehension, engagement levels, and learning progress while ensuring the experience remains positive and pressure-free.
Emotional Response Detection Introduce sentiment analysis to detect the emotional tone of a child’s responses. This will allow the system to better understand how the child reacts to stories and adjust storytelling style accordingly.
Adaptive Moral Reinforcement Engine Develop a more advanced adaptive engine where the AI dynamically adjusts future stories based on the child’s answers, engagement level, and understanding. This will strengthen moral learning through personalized storytelling paths.
Transforming the POC into a Real Product The long-term goal is to evolve AIM into a complete application that provides children with a fun, adaptive, and value-based storytelling experience that grows with them.
Built With
- angular.js
- bedrock
- java
- nova
- postgresql
- python
- spring
Log in or sign up for Devpost to join the conversation.