Why LLM Recommenders Will Be AI's Biggest Consumer App — Devansh Tandon, Meta

Why watch

Devansh Tandon, Meta's Recommendations Research lead, explains why LLM recommender systems will become AI's biggest consumer app. He outlines four S-curves (traditional recsys, LLM-inspired, LLM-native, agentic) and details the implementation recipe: tokenizing content with semantic IDs, bilingual model training, and post-training for ranking with interpretable reasoning. The talk demonstrates how Instagram achieved 30% more watch time and shows how feeds become steerable. For ML engineers and product managers building recommendation systems.