Design a scalable chatbot platform
Company: Atlassian
Role: Machine Learning Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
## Problem
Design a production chatbot platform that can answer user questions and hold multi-turn conversations for a product/company.
Assume the chatbot will be used in web and mobile apps and should be able to:
- Answer FAQs and product/how-to questions using internal documentation/knowledge bases.
- Support multi-turn conversations with conversation history (context).
- Escalate to a human agent when it cannot help.
- Be safe (avoid leaking sensitive info) and observable (monitoring/metrics).
## Deliverables
1. Clarify requirements (functional + non-functional), traffic/latency targets, and failure modes.
2. High-level architecture and key services/components.
3. Data/storage design (conversation state, knowledge sources).
4. Key APIs (ingress, inference, human handoff).
5. Reliability, scaling, and monitoring plan.
6. Security/privacy considerations and abuse prevention.
Quick Answer: This question evaluates a candidate's ability to design a scalable, production-grade chatbot platform that integrates ML-based inference, multi-turn conversational context management, knowledge retrieval from internal documentation, human handoff workflows, and security/privacy controls.