Design a ChatGPT-like conversational assistant
Company: Plaid
Role: Software Engineer
Category: System Design
Difficulty: easy
Interview Round: Technical Screen
Design a ChatGPT-like conversational assistant product.
Assume you need to support:
- Multi-turn chat with conversation history
- Streaming responses (token-by-token)
- High availability and low latency at peak traffic
- Basic safety controls (prompt injection awareness, toxicity filtering, PII handling)
- Optional tool use (e.g., calling internal search or a calculator)
Describe the high-level architecture, key components, data storage, scaling strategy, and how you would evaluate quality and safety in production.
Overview: This question evaluates a candidate's ability to design large-scale conversational AI systems, covering competencies in distributed system architecture, real-time streaming, stateful multi-turn session management, safety controls, and integration with external tools.
Read the full Plaid Software Engineer interview experience this question came from