Katkat
Back

Design trending livestream discovery

WhatnotMachine Learning Engineer · Technical Screen
MediumMachine Learning EngineerTechnical ScreenML System Design

Last updated

This question evaluates expertise in designing end-to-end ML-driven recommendation systems, including competencies in real-time feature engineering, candidate generation, model serving, business-rule integration, monitoring, experimentation, and operational trade-offs for live content.

Design a system for a live-commerce platform that surfaces trending livestreams to users.

Assume an ML model for scoring trendiness or relevance already exists and can be called as a service. Your task is to design the end-to-end ML system around that model.

Discuss:

  • the product goal and success metrics
  • what events and features you would collect
  • how to generate candidate livestreams
  • how to combine business rules, real-time signals, and the model score
  • the online and offline architecture
  • latency, freshness, and scaling requirements
  • how to prevent stale, low-quality, spammy, or ended streams from being recommended
  • experimentation, monitoring, and failure handling

Follow-up: How would you ensure that a stream shown in the trending feed is actually still live?