This was my second round system design interview at Meta. The question was about designing the search and ranking system for Marketplace.

I started with the high level: ingestion pipeline for listings, search index, ranking service. The interviewer quickly moved past the basics and wanted to talk about ranking signals.

We spent a good chunk of time on how to rank results. I proposed a two-phase ranking: a fast candidate retrieval using Elasticsearch with basic filters (location, category, price), followed by a ML re-ranking model that considers user behavior signals.

The follow up was about handling spam listings. Discussed content moderation pipeline with image classification and text analysis. Also talked about how to handle the cold start problem for new sellers with no reputation.

One thing that caught me off guard was the question about A/B testing the ranking model. Had to think on my feet about how to measure marketplace health metrics beyond just click through rate.

Got the offer. Total comp was better than expected honestly.