Ken Kennedy Lecturer

Dr. Omar Florez is an AI researcher working on the efficient training of open-source LLMs and the development of sovereign artificial intelligence in Latin America. He leads pre-training for LatamGPT, the first foundation model trained in the region: a 70B-parameter model trained on 300B tokens of regionally grounded data, built to reflect the linguistic and cultural diversity of its Spanish- and Portuguese-speaking communities.

He previously spent over a decade in Silicon Valley as a senior research scientist at Twitter Cortex, Intel Labs, and Capital One. His work spanned multilingual language modeling, large-scale distributed training, and dialogue systems, and included contributions to early multilingual and multimodal language models trained at Twitter.

He is a founding member of LatinX in AI, which supports the participation of Hispanic researchers at leading AI conferences worldwide. He has also contributed to public policy, including Peru's National AI Strategy through the Office of the Prime Minister of Peru.

At CENIA (the National Center for Artificial Intelligence of Chile), he leads AI initiatives focused on training foundation models with regional data and advancing globally competitive alternatives to existing LLMs through open research and the public release of model weights, training datasets, training recipes, and evaluation benchmarks. He has authored 20 publications and holds 10 patents in machine learning and artificial intelligence.