NVIDIA engineers use Codex with GPT-5.5 for production systems and research

NVIDIA engineers and researchers are using OpenAI's Codex, built on GPT-5.5 and running on NVIDIA GB200 and GB300 infrastructure, to ship production systems and automate machine learning research workflows. The coding agents team, led by senior software engineer Dennis Hannusch, uses Codex as a default tool for complex engineering tasks, evolving an internal platform from MVP to production-ready and building an internal podcast recording app in hours. Codex's autonomous capabilities include testing video and audio recording without human intervention. For research teams, AI researcher Shaunak Joshi reports that Codex automates the entire research loop: identifying areas, writing scripts, and running experiments on remote machines via SSH. Codex also performs machine translation, converting Python code to Rust for up to 20x efficiency gains. The tool has enabled a 10x speed improvement in end-to-end research workflows. Hannusch notes that Codex with GPT-5.5 is more autonomous and maintains context over long sessions. The article was published by OpenAI Academy on May 12, 2026.