Snap Engineering

Overview

Snap is the company behind Snapchat, where ephemeral video and Spotlight short-form content drive the central ML problems and the user base skews young and mobile-first, which puts unusual pressure on personalization quality. The blog's most distinctive content covers three areas: Bento, an internal ML platform that standardizes feature engineering, training, and serving for hundreds of models across the company, embedding-based retrieval for video recommendations on Spotlight that has to surface the right clip in milliseconds, and TPU-based training infrastructure for very large recommendation models. For interviews, 'Snap's Embedding-based Retrieval for video recommendation' is a useful reference for two-tower retrieval design, and 'Bento - Snap's ML Platform' anchors discussions of building an ML platform end to end.

Interview questions

• What can I learn from Snap's engineering blog?
• Which Snap posts are most useful for ML-platform interviews?
• Why is Snap's blog valuable for recommendation-retrieval discussions?

More Engineering Blogs topics

• Quora Engineering
• Razorpay Engineering
• Reddit Engineering
• Salesforce Engineering
• Shopify Engineering
• Slack Engineering
• Spotify Engineering
• Stripe Engineering
• Swiggy Engineering
• Tinder Engineering
• Twitch Engineering
• Twitter/X Engineering