Design a search-to-ads ranking pipeline
Company: Snapchat
Role: Machine Learning Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
## Prompt
Design a high-level **search + ads ranking** system for an app where a user issues a query and the product shows a mix of organic search results and sponsored ads.
### Requirements
- Return results within a strict latency budget (assume ~200 ms end-to-end).
- Ensure good user experience while meeting monetization goals.
- Support experimentation (A/B tests) and continuous model iteration.
### Discuss
1. The main components of the request flow (retrieval, filtering, ranking, mixing).
2. How ads ranking differs from organic ranking (auction/bids, pacing/budgets, policies).
3. How you would mix ads and organic results and what constraints you would enforce.
4. Key metrics to monitor and how you would debug regressions.
Quick Answer: This question evaluates knowledge of designing a low-latency search-plus-ads ranking pipeline and related competencies in retrieval, filtering, multi-stage ranking, ad auction mechanics, pacing/budget constraints, result mixing, experimentation, and monitoring.