Company

Criteo

9 cases from Criteo (Adtech, E-commerce), each digested from a public talk or engineering post and linked to the original. 3 were checked in an online A/B test; 3 report results in production.

Tasks: Retrieve and rank from millions of items, Rank search results, Learn user and item embeddings, Recommend from what the user just did, Experiment when users affect each other and 3 more. Methods: Two-tower model, Contrastive learning, Negative sampling, In-batch negatives, Relevance labeling, Embedding fine-tuning and 19 more. Tools: Ray, CLIP, PyTorch, ONNX, NVIDIA Triton, XGBoost and 7 more.

Tech blog: medium.com/criteo-labs