Functional requirements (FR)

• Add functional requirements here.

Non-functional requirements (NFR)

• Add non‑functional requirements here.

1. Notification Platform Architecture

The prompt

Design a multi-tenant, cloud-native notification platform (SaaS) that allows internal and external microservices to dispatch notifications across Push, SMS, and Email with strict rate limiting, multi-client isolation, and carrier redundancy.

Functional requirements

• Multi-Channel Dispatch: Ingest and route messages across Mobile Push (APNs, FCM), SMS (Twilio, Sinch), and Email (SendGrid, SES).
• Multi-Tenant SaaS Isolation: Support tenant accounts with dedicated usage quotas, API credentials, and per-tenant rate limits.
• Pluggable Carrier Architecture: Enable adding or updating third-party communication providers without downtime or codebase refactoring.
• Dynamic Template Management: Render localized templates with dynamic JSON variable interpolation.
• Delivery Receipts & Telemetry: Ingest carrier webhooks to track real-time delivery status (QUEUED, SENT, DELIVERED, FAILED).

Non-functional requirements

• High Availability: 99.99% service availability with automated multi-carrier failover during vendor outages.
• Low Latency: P0 transactional alerts (e.g., OTPs) dispatch to carriers in $p99 \le 3\text{ seconds}$.
• Throughput: Scale to 50,000 dispatches/second during marketing campaigns without degrading OTP channels.
• Deduplication: Content-hash idempotency prevents duplicate alerts within a 10-minute sliding window.

Architecture Diagram

2. Distributed Job Scheduler Architecture

The prompt

Design a distributed, highly available job scheduling engine capable of executing delayed tasks and recurring cron expressions across a distributed worker fleet with lease fencing and zombie worker recovery.

Functional requirements

• One-off & Cron Schedules: Schedule tasks with arbitrary delays ($T + \Delta t$) or standard 5-field cron syntax.
• Task Lease & Execution Tracking: Coordinate task assignment to workers with active heartbeat leases and execution states.
• Automated Retries: Exponential backoff and dead-letter queue (DLQ) support for failed jobs.
• Management API: Support task pausing, cancellation, and manual re-triggering.

Non-functional requirements

• Dispatch Accuracy: Scheduled jobs trigger within $\le 500\text{ ms}$ of target timestamp.
• Zero Task Loss: At-least-once execution guarantee with persistent commit logs before worker dispatch.
• Scalability: Support 100M scheduled tasks with automated worker cluster auto-scaling.

Architecture Diagram

Discussion prompts

• How do you isolate transactional P0 queues from bulk marketing P3 queues to eliminate head-of-line blocking?
• What health checks and error-rate thresholds determine when to trip a circuit breaker on a degraded SMS carrier?
• How do distributed leases and fencing tokens prevent a delayed zombie worker from executing an already reassigned job?
• What are the storage trade-offs between Redis Sorted Sets and partitioned PostgreSQL for scheduling queues?