See how AWS and Snorkel Flow integrate to build valuable enterprise LLMs faster than ever

This webinar provides a comprehensive overview of using Snorkel Flow and AWS to fine-tune large language models (LLMs), focusing on practical applications and real-world scenarios. Featuring experts from Snorkel and AWS, the session outlines the critical steps involved in the fine-tuning process and emphasizes the importance of high-quality training data. The discussion begins with an overview of the technology involved in building better LLM systems, including generative models, the components of retrieval augmented generation (RAG) systems, and fine-tuning all of them. The video also covers the methodologies for fine-tuning, including the evaluation of embedding models and foundation models. Practical demonstrations illustrate how to create a quality model, label training data, and execute fine-tuning jobs efficiently. Colin Toal, Principal, Business Development, AI/ML AWS, then offers a thorough overview of the AWS ecosystem, detailing the capabilities of services like Bedrock and SageMaker. The presenters explain how these tools can enhance the fine-tuning of LLMs, showcasing the integration with Snorkel Flow for data curation and model evaluation. Viewers will learn about the iterative nature of fine-tuning, understanding how to assess model performance and make necessary adjustments. The session is tailored for professionals in the field of machine learning who want to deepen their understanding of AWS capabilities in the context of LLM optimization. TIMESTAMPS: 00:00 Introduction and Speaker Introductions 00:58 Agenda Overview and Fine-Tuning Discussion 02:30 Snorkel's Role in Optimizing RAG Pipelines 06:30 Embedding Models and Fine-Tuning Techniques 16:35 Training Data and SME Involvement 24:15 AWS Integration and Generative AI Applications 26:44 Comprehensive Offerings for Machine Learning 27:12 Amazon Bedrock: Features and Benefits 27:36 Model Customization and Security in Bedrock 28:07 Diverse Model Options in Bedrock 29:35 Data Privacy and Compliance in Bedrock 30:51 Amazon SageMaker: Tools and Capabilities 31:57 Demo Introduction: Fine-Tuning Workflow 32:04 Setting Up the Fine-Tuning Challenge 33:48 Developing and Evaluating the Fine-Tuned Model 36:02 Annotation and Quality Model Development 39:46 Fine-Tuning Process and Results 48:49 Use Cases and Applications of LLMs 50:32 Q&A: Training Data and Fine-Tuning Strategies 55:55 Conclusion and Final Remarks #aws #largelanguagemodels #rag