Emerging Architectures for Large Language Model Applications

活动简介

Generative AI and Large Language Models have taken the world by storm. Applications like Bard, ChatGPT, Midjourney, and Dall.E crossed the proverbial chasm of technology adoption lifecycle and have entered some applications like content generation and summarization. However, there are inherent challenges for a lot of tasks that require a deeper understanding of trade-offs like latency, accuracy, and consistency of responses. Any serious applications of LLMs require an understanding of nuances in how LLMs work, embeddings, vector databases, retrieval augmented generation (RAG), orchestration frameworks, and more.

This introductory tutorial will introduce the audience to prevalent approaches to building a custom large language model application. We will present a canonical architecture for an LLM application and various available commercial and open-source tools and technologies available to build these applications.

By the end of the talk, you will understand:

- Typical architecture of an LLM application
- Processing of single query/task/inference job completed in in-context learning even when the prompt is longer than the context window of the foundation LLM
- Role of embeddings and vector databases like vector databases during in-context learning
- Role of orchestration frameworks like LangChain and LlamaIndex
- The need for a semantic cache when building applications at scale
- Challenges and pitfalls faced while building these applications to solve real-world problems

No prior background in Generative AI or LLMs is necessary to attend this talk.