Beyond the Model Zoo: Optimizing Foundation Models for Your Application

Salma Mayorquin, Co-Founder of Remyx AI, discusses how to optimize foundation models for applications. The talk covers parameter-efficient fine-tuning (PEFT) methods like LoRA, the importance of data synthesis and augmentation for creating high-quality datasets, and advanced LLM evaluation techniques for assessing model performance. Mayorquin also touches on deployment strategies using tools like Triton servers and emphasizes that achieving state-of-the-art model performance is now possible with fewer resources than in the past. She highlights the benefits of using custom evaluation criteria and leveraging generative AI for both data design and model assessment.