What Every LLM Developer Needs to Know About GPUs with Charles Frye
Join a live podcast recording with Charles Frye, Developer Advocate at Modal and expert in machine learning systems, in conversation with host Hugo Bowne-Anderson.
In this episode, we’ll discuss:
Why GPUs Are Essential for LLMs: Understanding the hardware features that make GPUs the backbone of AI workloads.
GPU Internals, Simplified: Key concepts like Streaming Multiprocessors (SMs), PTX, CUDA, and memory hierarchies, and how they enable scalable AI systems.
Practical Advice for Developers: When and how to use GPUs effectively in LLM workflows.
Scaling AI Workloads: Insights into building efficient pipelines to maximize performance on GPU hardware.
This episode is ideal for LLM developers, AI engineers, and anyone interested in understanding how GPUs power modern AI systems.
We’re diving into this topic now because Charles and the Modal team recently launched their GPU Glossary, a resource that demystifies GPU hardware and software, making it easier for developers to master the tools that power modern AI. Check it out!
About Charles Frye
Charles Frye is a Developer Advocate at Modal, where he works on tools and resources to simplify GPU programming for developers. He also teaches people about machine learning systems and programming through the Full Stack Deep Learning course and community. Previously, Charles worked at Weights & Biases, where he developed tools, technical documentation, and educational content for machine learning practitioners.
Charles holds a PhD in Computational Neuroscience from UC Berkeley, where he focused on artificial neural networks, supported by the NSF Graduate Research Fellowship. His technical interests span neural networks, computational Bayesian methods, and category theory for programming.