The brief
Sandesh is a multi-channel notification delivery system built during an engineering internship for an AI consulting and training company.
The company needed a fast and reliable way to deliver notifications across push notifications, email, and SMS while supporting prioritisation, retries, scheduling, templating, and failure handling.
Rather than introducing a heavyweight distributed messaging platform that was unnecessary for the company’s scale, we designed a lightweight priority-queue-based architecture tailored to the actual workload.
Under the measured operating conditions, the system achieved sub-300 ms push notification delivery with 95% confidence and approximately 98% email delivery reliability.
One delivery layer for multiple channels
Sandesh provides a unified delivery layer across push notifications, email, and SMS.
Applications can submit notifications without managing the underlying delivery mechanics themselves. Sandesh handles queueing, prioritisation, scheduling, provider interaction, retries, and delivery state.
Priority-based delivery
Not every notification has the same urgency.
A critical system alert should not necessarily wait behind routine notifications. Sandesh therefore uses priority-aware queuing to determine the order in which notifications are processed.
Higher-priority messages can move through the delivery pipeline ahead of lower-priority work, allowing the system to respond quickly to time-sensitive events without requiring heavyweight messaging infrastructure.
Reliable delivery through controlled retries
External notification providers can fail because of temporary network problems, provider errors, rate limits, invalid destinations, or transient service outages.
Sandesh treats delivery as a process rather than a single attempt.
When an appropriate failure occurs, the notification can be placed back into the delivery workflow according to the retry policy. This prevents transient failures from immediately becoming permanent delivery failures while avoiding uncontrolled retry loops.
Scheduling and templating
Sandesh supports scheduled delivery, allowing applications to specify when a notification should become eligible for processing.
It also supports reusable notification templates so applications can maintain consistent notification structures while dynamically inserting information for individual messages.
Reliability without unnecessary infrastructure
One of the key architectural decisions was deliberately not introducing Kafka.
The organisation’s notification volume did not justify the operational complexity of a large distributed event-streaming platform. Instead, the system was designed around the actual scale and requirements of the business.
The priority-queue architecture provided ordering, retries, failure handling, scheduling, and controlled delivery without unnecessary infrastructure complexity.
Performance that mattered
Performance was treated as a core requirement.
Under the measured operating conditions, Sandesh achieved:
Sub-300 ms push notification delivery — 95% confidence
and approximately:
98% email delivery reliability
These measurements helped validate that a lightweight architecture could meet the company’s requirements without sacrificing responsiveness or reliability.
A reusable delivery layer
Sandesh separates notification creation from delivery.
Applications submit messages while the delivery platform manages queueing, prioritisation, scheduling, provider interaction, retries, and delivery outcomes.
This creates a reusable foundation that can support multiple products and internal systems as notification requirements grow.
The engineering principle
Use the simplest architecture that reliably solves the actual problem.
Instead of reaching immediately for a large distributed messaging system, the project focused on understanding the workload, identifying the required guarantees, and building only the infrastructure necessary to provide them.
The result was a fast, reliable, and operationally lightweight notification platform capable of supporting multiple delivery channels and real-world failure conditions.
Project highlights
- Multi-channel push, email, and SMS delivery
- Priority-based notification queues
- Controlled retry and failure handling
- Scheduled notifications
- Reusable notification templates
- Sub-300 ms push delivery at 95% confidence
- Approximately 98% email delivery reliability
- Lightweight alternative to heavyweight messaging infrastructure
- Firebase-based push delivery
- MongoDB-backed application infrastructure
- React-based management interface