DoorDash·Software Engineer·Onsite - System Design / Architecture·Senior

Senior

Jun 2026

Summary

System design round at DoorDash for a software engineering role. The prompt was a full-blown alert notification platform, the kind where you think you know where to start and then realize thirty minutes in that you've barely scratched the surface.

Questions Asked (1)

Design a multi-tenant alert notification system for operational incidents, supporting millions of events per day, sub-30-second delivery, at-least-once guarantees, deduplication, escalation chains, on-call schedules, and failover across notification providers.

System DesignAPI & IntegrationsTechnical Trade-offs

Author's notes

I spent way too long on the data model upfront and the interviewer had to nudge me toward the actual delivery pipeline.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying requirements and constraints, then design a high-level architecture that separates event ingestion, processing, and delivery. Focus on trade-offs for scalability, reliability, and multi-tenancy, and dive into critical components like deduplication and escalation.

Pro tip: Emphasize idempotency and at-least-once semantics with deduplication keys to handle duplicate events gracefully. Also, discuss how to isolate tenants to prevent noisy neighbor issues and ensure fair resource allocation.

1. Clarify Requirements and Scope

Ask questions to understand scale, latency, delivery guarantees, and multi-tenancy needs. Confirm assumptions about event volume, tenant count, and notification channels.

2. High-Level Architecture

Outline components: ingestion API, message queue, processing workers, deduplication store, notification dispatcher, and provider integrations. Explain data flow from event to delivery.

3. Deep Dive into Key Components

Detail deduplication using idempotency keys and TTL caches, escalation chains with state machines, on-call schedules with calendar integration, and provider failover with circuit breakers.

4. Scalability and Reliability

Discuss partitioning, sharding, and autoscaling for millions of events. Explain how to achieve sub-30-second delivery with at-least-once guarantees and handle failures.

5. Trade-offs and Multi-Tenancy

Analyze trade-offs between consistency and availability, cost vs. latency, and tenant isolation strategies. Summarize how the design meets requirements.

Key Points to Mention

Use of idempotency keys and deduplication windows to ensure exactly-once processing effect.
Escalation policies with time-based state machines and integration with on-call schedules (e.g., PagerDuty).
Multi-tenant isolation via per-tenant queues, rate limiting, and resource quotas to prevent noisy neighbors.
Provider failover with health checks, circuit breakers, and fallback to secondary providers.
At-least-once delivery with acknowledgment and retry mechanisms, and dead-letter queues for poison messages.
Partitioning strategies (e.g., by tenant ID or event type) to scale horizontally and meet latency SLAs.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.