← Head back to all of our AI Engineer World's Fair recaps
Decoding Mistral AI's Large Language Models
Devendra Chaplot @dchaplot / Mistral
Watch it on YouTube | AI.Engineer Talk Details
This talk is about how Mistral AI trains and uses large language models. The speaker explains the steps involved in creating these AI models and shares some tips on when to use open-source models versus commercial ones.
No more bad outputs with structured generation
Rémi Louf @remilouf / .txt (Outlines)
Watch it on YouTube | AI.Engineer Talk Details
This talk introduces Outlines, an open-source library that allows developers to generate structured outputs from language models. The speaker argues that using structured generation solves common problems with language models and improves their performance, making them more reliable and efficient for various tasks.
⭐ Building SOTA Open Weights Tool Use: The Command R Family
Sandra Kublik @itsSandraKublik / Cohere
Watch it on YouTube | AI.Engineer Talk Details
This talk is about Cohere's Command R family of AI models. Sandra Kublik explains how these models are designed to be really good at using tools and finding information, while being smaller and cheaper to use than some bigger AI models.
Read our Deep Dive on this talk as well.
Training Albatross: An Expert Finance LLM
Leo Pekelis @LeoPekelis / Gradient
Mark Kim-Huang @markatgradient / Gradient
Watch it on YouTube | AI.Engineer Talk Details
This talk is about how Gradient created a specialized AI model for finance called Albatross. The speaker explains how they trained this model to understand financial topics better than general-purpose AI models, and also how they made it able to work with much longer pieces of text.
Fixing bugs in Gemma, Llama & Phi-3
Daniel Han @danielhanchen / Unsloth
Watch it on YouTube | AI.Engineer Talk Details
This talk is about fixing bugs in open-source AI models like Llama 3, Gemma, and Phi-3. The speaker shares several issues they found in these models and explains how to fix them, focusing mainly on problems that can occur when fine-tuning Llama 3.
⭐ Everything you need to know about Finetuning and Merging LLMs
Maxime Labonne @maximelabonne / Liquid AI
Watch it on YouTube | AI.Engineer Talk Details
This talk covers everything you need to know about fine-tuning and merging large language models (LLMs). Maxime Labonne explains different techniques for improving LLMs, like fine-tuning them on specific data and combining multiple models, to make them work better for different tasks.
Read our Deep Dive on this talk as well.