System Design: Scalable Interview Question Bank Platform

Context

Design a multi-tenant SaaS platform for organizations to create and manage an interview question bank used during live technical interviews. The system must support authoring, collaborative review, tagging, difficulty calibration, fast search/browse, strict versioning, and a menu-style selection flow used by interviewers in real time.

Assume a target scale such as:

• 5,000 organizations, 2 million total questions, 50 million versions, 500 million events (views/uses), 200k monthly active users (authors/interviewers), 20k peak RPS reads, 1k peak RPS writes.
• P95 API latency targets: reads ≤150 ms, writes ≤300 ms, search ≤300 ms, selection flow ≤150 ms.

Requirements

1. Core features
   • Authoring with rich text and attachments; collaborative editing and comments.
   • Tagging (topics, skills, languages), difficulty calibration, and versioning (immutable versions, draft → review → publish workflow).
   • Search and browse with filters/facets; semantic and keyword search; suggestions.
   • Menu-style selection flow for live interviews (guided, filterable, and diverse question menus).
2. Technical specifications to provide
   • High-level architecture and data/storage choices.
   • API endpoints covering CRUD, versioning/publish, tagging, search, selection, calibration, analytics.
   • Data models and indexing strategy (keyword + semantic; facets; incremental updates).
   • Recommendation/ranking approach for the menu flow (scoring and diversification).
   • Multi-region deployment and data replication strategy.
   • Caching layers and invalidation.
   • Consistency vs availability trade-offs by subsystem.
   • Rate limits and quotas.
   • Monitoring, alerting, and SLOs.
   • Cost considerations and optimizations.
   • Permissions/roles and abuse prevention (moderation, audit, anti-scraping, usage caps).
   • Offline editing and sync/conflict resolution.

State assumptions where needed and justify design choices.