AWS Builder Center
Weekend Annoying Task Challenge: RemindMe AI – An AI-Powered Voice Reminder Application

Weekend Annoying Task Challenge: RemindMe AI – An AI-Powered Voice Reminder Application

Built a serverless AI-powered reminder application using Flutter and AWS services including Amazon Bedrock, Amazon Polly, AWS Lambda, Amazon S3, and DynamoDB. This article explains the architecture, development process, challenges, and lessons learned while creating RemindMe AI.

Introduction

In today's fast-paced world, remembering important tasks can be challenging. Traditional reminder applications are useful, but they often require users to manually enter task details, dates, and times, making the process repetitive and time-consuming. I wanted to build a smarter reminder application that allows users to create reminders naturally by simply speaking, without navigating multiple forms or menus.
To solve this, I developed RemindMe AI, an AI-powered voice reminder application built using Flutter and AWS Serverless services. The application uses Amazon Bedrock to understand natural language, Amazon Polly to generate realistic voice reminders, Amazon S3 to securely store generated audio, Amazon DynamoDB to persist reminder metadata, and AWS Lambda to orchestrate the backend workflow.
Unlike many reminder applications that rely on cloud scheduling, RemindMe AI downloads the generated audio to the device after creation and uses OS-level scheduling through Flutter's local notification system. This allows reminders to trigger and play even when the device has no internet connection.
This project demonstrates how AI, serverless computing, and mobile technologies can work together to create an intelligent, scalable, and reliable productivity application.



Solution Overview

RemindMe AI provides a conversational reminder experience.
Instead of manually filling out forms, users simply say:
"Remind me to pay the electricity bill tomorrow at 9 PM."
The application automatically:
  • Converts speech into text.
  • Sends the transcript to Amazon Bedrock.
  • Extracts structured reminder information.
  • Generates a natural voice reminder using Amazon Polly.
  • Stores reminder metadata in Amazon DynamoDB.
  • Uploads the generated audio to Amazon S3.
  • Downloads the audio file securely to the user's device.
  • Stores the audio locally for offline playback.
  • Schedules the reminder using the device's operating system.
  • Displays the reminder immediately inside the application.
Once a reminder has been created successfully, it no longer depends on an internet connection. At the scheduled time, the operating system triggers the reminder, and the locally stored Polly-generated audio is played automatically.

Architecture

RemindMe AI follows Clean Architecture on the Flutter frontend and a fully serverless backend on AWS.
The backend is responsible for AI processing, reminder creation, audio generation, and secure storage. Once a reminder is created, the application downloads the generated audio from Amazon S3 using a secure pre-signed URL and stores it locally on the device.
Unlike traditional cloud-based reminder systems, the reminder scheduling itself happens entirely on the mobile device. Flutter schedules the reminder using Android's OS-level alarm and notification mechanisms, allowing reminders to trigger even when the application is closed or when the device is offline.
This hybrid architecture combines the scalability of AWS with the reliability of local device scheduling.




How It Works

When the user taps the microphone button, Flutter uses the device's speech recognition engine to convert speech into text.
The transcript is sent to an AWS Lambda function through Amazon API Gateway.
Amazon Bedrock analyzes the sentence and extracts structured reminder information, including:
  • Task
  • Date
  • Time
  • Priority
  • Repeat interval
If the confidence score is high, the reminder is created automatically. Otherwise, the application displays an AI confirmation screen where users can review and edit the extracted information.
After confirmation, another Lambda function invokes Amazon Polly to generate a natural-sounding MP3 reminder.
The generated audio is uploaded to Amazon S3, while reminder metadata is stored in Amazon DynamoDB.
After the reminder is successfully created, the Flutter application requests a secure pre-signed URL from the backend, downloads the MP3 file from Amazon S3, and stores it locally on the device.
Finally, Flutter schedules an exact local notification using the device's operating system. At the scheduled time, the operating system wakes the application if necessary, triggers the notification, and plays the locally cached audio file. Because the audio already exists on the device, reminders continue to work even when there is no network connectivity.



AWS Services Used

This project integrates several AWS services, each serving a specific purpose.

Amazon API Gateway:

Provides secure REST endpoints for the Flutter application.

AWS Lambda:

Executes all backend business logic without managing servers.

Amazon Bedrock:

Analyzes natural language and extracts reminder information using AI.

Amazon Polly:

Generates realistic neural speech for every reminder.

Amazon DynamoDB:

Stores reminder metadata in a scalable NoSQL database.

Amazon S3:

Stores generated MP3 files securely.

Amazon CloudWatch:

Monitors Lambda execution, logs API requests, and helps diagnose issues during development.



Development Journey

The development process involved building both the frontend and backend simultaneously.
I started by designing the Flutter application using Clean Architecture with Riverpod for state management. Once the UI and networking layers were ready, I implemented the backend using AWS SAM, Lambda, API Gateway, and DynamoDB.
After the basic CRUD functionality was working, I integrated Amazon Bedrock to understand natural language reminders and Amazon Polly to generate spoken reminders.
Finally, I focused on improving the overall user experience by introducing AI confirmation screens, secure audio streaming using S3 pre-signed URLs, and better error handling throughout the application.









Challenges and Solutions

One of the biggest challenges was handling streaming audio generated by Amazon Polly.
Initially, uploading Polly's audio stream directly into Amazon S3 caused header validation errors. After debugging the issue using Amazon CloudWatch, I modified the upload process to properly handle the streaming response before sending it to S3.
Another challenge involved debugging communication between Flutter and AWS APIs. CloudWatch logs proved invaluable for tracing requests, validating payloads, and identifying missing headers or incorrect request formats.
Integrating multiple AWS services while maintaining a clean, modular architecture was another important learning experience throughout the project.



What I Learned

This project helped me gain hands-on experience with several AWS services and cloud-native development practices.
Some of my key learnings include:
  • Building serverless applications using AWS Lambda.
  • Designing scalable APIs with Amazon API Gateway.
  • Integrating AI capabilities using Amazon Bedrock.
  • Generating realistic speech using Amazon Polly.
  • Managing secure file storage with Amazon S3.
  • Designing efficient DynamoDB schemas.
  • Using CloudWatch for debugging and monitoring.
  • Applying Clean Architecture and Riverpod in Flutter applications.
Beyond technical implementation, this project strengthened my understanding of designing scalable, maintainable, and production-ready cloud applications.



Future Improvements

Although the core functionality is complete, I plan to continue enhancing RemindMe AI with additional features:
  • Background reminder execution even when the application is closed.
  • Exact Android alarm scheduling.
  • Offline reminder synchronization.
  • Reminder history.
  • Snooze functionality.
  • Calendar view.
  • Multiple Amazon Polly voice selections.
  • Cross-device synchronization.
  • User authentication.
  • AI-powered recurring reminder suggestions.
These enhancements will move the application closer to a production-ready personal assistant.



GitHub Repository

GitHub Repository

Live Demo

Android APK Download:
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article