GenAI Cracow #4

PRESENTATIONS

​Vulnerabilities of LLMs

During my presentation, I will discuss attacks on large language models (LLMs) and their classification. I will cover the sources of some vulnerabilities, share cases that have been reported in the media, and we might even have time to consider how we can mitigate these issues.

​RAGOps: Building Advanced RAG Systems with Open-Source Large Language Models

This presentation addresses the growing challenge enterprises face in utilizing open-source Large Language Models (LLMs) for developing advanced Retrieval-Augmented Generation (RAG) applications in complex business environments. The methodology combines recent advancements in open-source LLMs with optimized RAG strategies, involving the integration of diverse data sources, enhanced query mechanisms, and fine-tuned model responses. The approach emphasizes a symbiotic relationship between retrieval and generation phases, enriched by expert feedback, to create an efficient RAG pipeline. Results demonstrate superior performance over baseline benchmarks, showing improved precision, contextual understanding, scalability, and adaptability to enterprise requirements. The talk provides industry professionals with practical insights into deploying LLMs to enhance customer service, market intelligence, and content creation, thereby fostering operational excellence and innovation.

​Refactor legacy code using LLMs - case study

More details soon