Dream11 Unveils ‘Lumos’ Foundational AI Model to Scale Personalization for 250M+ Users

Dream11 has shared an update. The company detailed the development of “Lumos,” an internal foundational intelligence model that consolidates more than 50 task-specific personalization systems into a single architecture. According to the post, Lumos is designed to scale across Dream11’s user base of over 250 million fantasy sports players and is trained on approximately 50 petabytes of data, incorporating elements inspired by large language model (LLM) architectures to better understand user behavior across different sports, formats, and contexts.

For investors, this initiative signals a strategic push toward deeper data-driven personalization and operational efficiency. Replacing numerous task-based models with one foundational system could reduce technical complexity and maintenance costs over time, while improving recommendation quality and user engagement metrics such as session length, retention, and frequency of play. Enhanced personalization can also support more targeted monetization, including contest participation, cross-sell of other products within the Dream Sports ecosystem, and potentially higher lifetime value per user.

At an industry level, the move underscores increasing adoption of advanced AI and foundation models in sports-tech and fantasy gaming, where personalization and real-time relevance are competitive differentiators. If successful, Lumos could strengthen Dream11’s defensibility against both domestic and international rivals by making its product experience more adaptive and harder to replicate. However, the financial impact will depend on execution quality, infrastructure costs of training and serving such a large model, and the company’s ability to translate engagement gains into sustainable revenue growth while managing regulatory and data-privacy considerations in its key markets.