How NVIDIA engineers and researchers build with Codex
NVIDIA leverages the autonomous capabilities of Codex, powered by GPT-5.5, to accelerate complex engineering tasks, research workflows, and production system development.Summary
NVIDIA engineers and researchers are utilizing Codex, built on GPT-5.5 and running on NVIDIA GB200/GB300 infrastructure, as a primary tool for autonomous software development and machine learning research. The model excels at long-session tasks, identifying bugs, and managing end-to-end workflows that were previously difficult to automate.
For engineering teams, Codex has proven effective in scaling internal platforms and rapidly developing production-ready applications, such as a custom podcast recording tool. Meanwhile, research teams use it as an autonomous partner to analyze academic papers, manage machine learning experiments on remote infrastructure, and modernize legacy codebases by translating Python into more efficient languages like Rust.
(Source：OpenAI)