Criteo x NVIDIA AI Meetup: LLM-Powered Recommendations at Scale

NVIDIA and Criteo take LLM-powered recommendations from prototype to production in one evening. Criteo's Zakaria Ouzarf and Marius Alonso explain how they built a product relevance classifier on open-weight models that runs entirely on-premises, covering fine-tuning, distillation, quantization and inference tuning, with the dead ends and the hard numbers. NVIDIA's Marek Pikulski covers the inference stack behind generative recommendation, including beam search and multi-item scoring in TensorRT LLM. Doors 18:00, talks from 18:30, a panel, then food and drinks. Free, but registration needs host approval and the exact address is shared only once you are accepted. It is in Paris. Why go: a rare chance to hear production numbers on running LLM recommendation systems on your own hardware. Best for ML and inference engineers.