Ranking @ Snap Inc - Content Retrieval & Recommendation System
Founding member of Snapchat short-form video recommendation product (a.k.a Spotlight). Working on large-scale ML recommendation systems, including multi-stage ranking, retrieval, training data, feature engineering, model inference, etc. Help Snapchatters create and consume content to make them feel closer to their friends, get entertained, and learn about the world.
Areas of focus
• Creator content recommendation, Creator signal, and Creator success
• Inverted-Index-based Retrieval, Graph-based Retrieval, Embedding-based Retrieval
• Spotlight Video Storage, Retrieval & Ranking, and Personalization
• Machine Learning & Ranking Product (Kubernetes, Microservice, Distributed System)
Projects:
• Creator content recommendation strategies: Inverted-Index-based Retrieval & similar item/user retrieval
• Graph-based Retrieval: Heterogeneous graph, collaborative filtering, relationship & interaction data
• Embedding-based Retrieval for cold-start exploration: candidate indexing, feature sourcing, Two-Tower model training, Airflow orchestration, embedding generation, online embedding inference, candidate retrieval service
• Creator signal generation and delivery system for 20+ key ranking and product services.
• Content Retrieval Service for the recommendation system: utilize multiple key micro-service components and distributed systems with Kubernetes.
• Spotlight story lifecycle (creation, amendment, deletion, storage, and management)
