Most applications send notifications using fixed schedules or generic delivery rules.
The challenge is that users have different engagement patterns. A notification sent at the wrong time is more likely to be ignored, contributing to notification fatigue and lower engagement.
Recently, I came across an interesting project by Yadab Sutradhar called Smart Notification Routing Engine, which tackles this problem using machine learning to predict the optimal notification delivery time for each user.
GitHub Repository:
https://github.com/Yadab-Sd/smart-notification-routing-engine
Key Features
- ML-driven send-time optimization
- Real-time prediction using Amazon SageMaker
- Event-driven architecture
- Multi-channel notification support
- AWS serverless infrastructure
- Automated model training pipelines
Architecture
The project leverages:
- AWS Lambda
- Amazon SageMaker
- Kinesis Data Streams
- EventBridge Scheduler
- DynamoDB
- Apache Spark
- AWS CDK
How It Works
- User interaction events are collected through the ingestion layer.
- Spark ETL jobs transform raw event data into machine learning features.
- SageMaker trains and deploys prediction models.
- A decision service predicts the best notification delivery time.
- EventBridge automatically schedules notifications for delivery.
Why It's Interesting
Instead of relying on static notification schedules, this system aims to personalize delivery timing based on user behavior, helping improve engagement while reducing unnecessary notification overload.
If you're interested in MLOps, event-driven systems, AWS serverless architectures, or recommendation and personalization systems, this project is worth exploring.
Repository:
https://github.com/Yadab-Sd/smart-notification-routing-engine
Would love to hear what others think about this approach to intelligent notification delivery.
Top comments (0)