Design Product Search Autocomplete

1. Problem Statement

2. Architecture Discussion Map

Use this as one discussion aid, not a single correct answer. Your design should follow from the requirements, scale, and trade-offs you establish.

graph LR
    A["Design Product Search Autocomplete"]
    A --> F1["Candidate generation versus model ranking"]
    A --> F2["Serve-time personalization, caching, and fallbacks"]
    A --> F3["Trending-query pipeline with minute-level freshness"]
    A --> F4["Marketplaces, locales, and cold start"]
    A --> F5["Privacy, safety, and abuse filtering"]

3. Key Focus Areas

• 1
  Candidate generation versus model ranking
• 2
  Serve-time personalization, caching, and fallbacks
• 3
  Trending-query pipeline with minute-level freshness
• 4
  Marketplaces, locales, and cold start
• 5
  Privacy, safety, and abuse filtering

4. What the Interviewer Looks For

The interviewer probes these areas, and the report grades the interview against them.

• Requirements, latency budget, and estimation
• Candidate generation versus ranking
• Serve-time personalization and caching
• Trending and fresh-query pipeline
• Marketplaces, locales, and cold start
• Privacy, safety, and abuse

5. What Strong Candidates Should Demonstrate

• Split autocomplete into candidate generation and model ranking, and size each stage.
• Blend per-shopper signals into ranking within a strict latency budget, with fallbacks.
• Decide what stays cacheable once results are personalized.
• Surface trending queries within a minute while keeping manipulation and private searches out of suggestions.