Jeremy Berchtold

Jeremy Berchtold is a senior software performance engineer at NVIDIA developing Transformer Engine, a library for accelerating large-scale transformer model training and inference. His work focuses on Transformer Engine JAX integrations and optimized implementations for Mixture of Experts, quantization, attention, and other performance-critical model components. Prior to Transformer Engine, he contributed to the autonomous vehicle stack at NVIDIA, focusing on complex intersection and junction handling.
Avatar photo

Posts by Jeremy Berchtold

MLOps

Accelerating Dropless MoE Training in JAX with NVIDIA Transformer Engine

Mixture of experts (MoE) has become one of the defining architectural trends in large-scale AI model training. DeepSeek, Qwen, and Mixtral are examples of MoE... 12 MIN READ