Luis Lastras

Luis is Director for Language and Multimodal Technologies, part of the AI Foundations team at the IBM T.J. Watson Research Center. The contributions of his department are directed towards IBM’s flagship enterprise AI Granite model efforts, and range from core contributions to tools, RAG, reasoning, multilingualism and synthetic data generation frameworks to separate model families including embeddings, vision and speech. His department also is working on new architectures for modular LLM constructions as well innovations in the intersection model architecture and runtime software. Trained as an information theorist, Luis has a keen interest in relying on solid theoretical foundations; past contributions include the main memory error correcting codes that IBM’s flagship Z and POWER servers used for multiple generations and has recently co-developed a new theory for the incorporation of mathematical logic in information theory and is additionally working on the information theoretic foundations of modular AI models. Luis obtained his Bachelor in Electrical Engineering from the Universidad Autonoma de San Luis Potosi, Mexico, and his MSc and PhD degrees from Cornell University in 1998 and 2000, respectively.