Abstract: Generalist Large Language Models have been implemented by some major tech players and made available to the public through subscription-based API. But these models come with risks, such as lack of transparency about the data used for their training, bias, impossibility to explain the generated results, incorrectness, lack of knowledge about specific business domains, intellectual property/sensitive data sharing, etc. which cannot be addressed by final users because of the closed source nature of these models and the service that expose them. What if you need a LLM specialized in some tasks related to your corporate domain area? During this talk you will learn more about Transformer-based models and some best practice to optimize domain specific fully Open Source LLMs to be deployed and used in private managed environments having computational power constraints. Attendees will learn about the critical importance of the Engineering more than Data Science behind LLMs management. A practical and real-life example related to a LLM that can generate working Python programming language code starting from natural language would be presented.
Bio: Guglielmo is a Biomedical Engineer with an extensive background in Software Engineering and Data Science applied to different contexts, such as Biotech Manufacturing, Healthcare and DevOps, just to mention the latest, and a lifelong learner. As part of the Manufacturing IT Advanced Mathematics and Modelling Data Science Team he is currently busy unlocking business value through Deep Learning projects, mostly in Computer Vision (not restricted to this field by the way). He has been recognized as DataOps Champion at the Streamsets DataOps Summit 2019 and awarded as one of the Top 50 Tech Visionaries at the 2019 Dubai Intercon Conference.
He is also an international speaker and author of the following book: Hands-on Deep Learning with Apache Spark @Packt https://www.packtpub.com/big-data-and-business-intelligence/hands-deep-learning-apache-spark.