Speakers
Fabian Klemm
Nithish Raghunandanan
Schedule
Location
As Python enthusiasts, we're all in a day-to-day struggle to explain ourselves: Why would you use Python? It's not even compiled, just dumb line-by-line interpretation! Can't you just do this in Rust? This talk aims to provide us with the holy grail of arguments in those discussions: We adapt (C)Python, so it has the best 'Hello, World!' of all! The talk will fail miserably in doing so. However, it should give some rudimentary insights into the fundamentals of Python: What happens to my code during execution? What is Python Byte-Code? What does the Python Virtual Machine do?
We have seen how AI tools like ChatGPT & Github Copilot hallucinate with some really convincing answers. In this talk, we can learn how to reduce these hallucinations by using a technique of combining AI chatbots with authoritative knowledge using open-source libraries like LangChain & LlamaIndex with the help of demos. The talk will offer practical tips on how to use AI with vector search and vector embedding to make searches more meaningful, intuitive, and tailored to user needs.
Retrieval Augmented Generation (RAG) has emerged as a popular design pattern for LLM applications, revolutionizing industry use cases. With a dynamic model and backend landscape, evaluating the answer quality of RAG pipelines is crucial in order to address issues like hallucination or irrelevance. This talk will guide you through the evaluation process and its current challenges, delving into key drivers, observability and metrics. For visualization, we will have a look at Phoenix (https://phoenix.arize.com/), an open-source library to facilitate tracing and evaluation.