DataHour: MemeGPT: Fine-Tuning LLMs to Generate Memes

In this DataHour session, the presenter will guide you in creating a Large Language Model specifically designed for meme generation. You'll learn the process of constructing your personalized Large Language Model through fine-tuning an open-source LLM, leveraging the power of NVIDIA HGX™ H100 and Habana Gaudi2 accelerators. Additionally, we'll explore the foundational aspects of LLMs, delving into linguistic complexities associated with humor, particularly focusing on memes.

Stefan is a seasoned professional with a diverse background in AI, product management, and data analytics. challenging IT projects in regulated environments with stakeholders spread across multiple time zones. Notably, he played a pivotal role in fostering a positive culture at Genesis Cloud, which ranked #6 among Top German Startup Employers in "Technology (AI, Cloud, BigData)" in 2022.With a rich portfolio of past projects, Stefan has demonstrated his proficiency in AI and machine learning. His accomplishments include the development of a Transformer-based time-series forecasting model for cryptocurrency pairs, a Large Language Model (LLM) for generating memes from the largest meme database, and an Explainable AI (XAI) model predicting credit rating transitions. His expertise extends to optimizing cloud GPU performance for ML model training, sentiment analysis, named entity recognition, and the creation of ML models for credit scoring and Cash Flow Available for Debt Service (CFADS).