Deep Learning with H2O.ai

Additional Speakers:
Philipp Singer, Senior Principal Data Scientist | Kaggle Grandmaster
Martin Barus, Data Scientist, Kaggle Master

Kaggle competitions in video sports analytics

In this talk we will explore the winning solutions from the recent Kaggle competitions in video sports analytics. In particular, soccer videos from the Bundesliga and American football videos from the NFL. We will examine the strengths and weaknesses of various architectures, namely 2D, 2.5D and 3D Convolutional Neural Networks (CNNs). The talk will be concluded with a demo of H2O Hydrogen Torch — a no-code Deep Learning platform, developed by Kaggle grandmasters.

Creating your own LLM

In this talk we will start with the general overview of Large Language Models (LLMs) and explore their training procedures. Then we will talk about the open-source h2oGPT LLM developed by H2O.ai. We will cover the technical details and make a demo of H2O LLM Studio: https://github.com/h2oai/h2o-llmstudio. It is an open-source tool that can be used for fine-tuning LLMs.