Teaching AI to Code in Every Language with NVIDIA NeMo

“Teaching AI to Code in Every Language with NVIDIA NeMo” is a hands-on training lab for machine learning practitioners, data scientists, and developers who want to build state-of-the-art coding language models on NVIDIA’s platform. Participants will learn how to use NVIDIA NeMo and NVIDIA GPU-accelerated infrastructure to train multilingual LLMs for code, from dataset preparation and cleaning to large-scale distributed training and optimization. The session will cover key challenges of multilingual code modeling, such as balancing languages, handling noisy repositories, and designing evaluation strategies that reflect real-world developer workflows. Attendees will implement practical techniques for scaling training across large compute environments, apply best practices for performance tuning, and explore how these models can power developer assistants, code review tools, and enterprise codebase management solutions. By the end of the lab, participants will have a clear roadmap to experiment with, train, and deploy advanced multilingual coding LLMs using NVIDIA’s end-to-end ecosystem.