At NVIDIA, engineers are using Codex as their default tool for complex engineering work, and to run end-to-end machine learning experiments. Codex, built on GPT‑5.5 and running in production on NVIDIA GB200 and GB300 infrastructure, can handle much longer, more autonomous sessions — going beyond execution to surface issues and ideas that weren't part of the original prompt.
“Codex is our go-to tool for complex engineering tasks, and with GPT-5.5, it surfaces bugs and gaps in my program that other models weren’t able to find.”
—Dennis Hannusch, Senior Software Engineer
Building and shipping production systems
NVIDIA’s coding agents team helps engineers across the company adopt and use AI tools effectively in real-world development workflows. Codex with GPT‑5.5 has become their go-to tool for complex engineering tasks.
“I’ve personally found Codex with GPT‑5.5 to be way more autonomous, with much less handholding,” explains Dennis Hannusch, a senior software engineer on the agents team. “I’m able to go for long sessions with multiple compactions and find that it still performs with top accuracy and manages to keep the work in context. And it’s great at tactically selecting the right tools as well as the right skills.”
Hannusch has already used Codex to evolve an internal platform from an MVP into a production-ready system, improving scalability and reliability along the way, something that had proven difficult with earlier models.
The team has also built an internal podcast recording app, similar to Riverside, spun up in just hours using Codex. “Given our privacy constraints, it would have taken us weeks to procure software,” Hannusch explains.
Using the Codex desktop app with computer interaction, the system was also able to test the video and audio recording functionality as it was built. “I didn’t have to do anything—it was built and tested completely autonomously,” he says. “Codex has completely changed the threshold for what’s worth building.”
“It’s been a 10x speed improvement just in terms of running experiments, because it’s able to handle the whole end-to-end machine learning research workflow.”
—Shaunak Joshi, AI researcher
Running full research workflows
For NVIDIA’s research teams, Codex has largely automated the research loop: from identifying research areas, to writing scripts for machine learning experiments, to running the experiments on remote machines.
“GPT‑5.5 has been a massive unlock as a creative partner, especially when it comes to knowledge work,” says Shaunak Joshi, an AI researcher at NVIDIA.
The team uses Codex as a research agent, pointing the model at a large corpus of relevant papers in areas like reinforcement learning. “GPT‑5.5 seems to be much more creative compared to competitors,” Joshi says. “It helped me trace snippets of evidence throughout the entire chain, and suggested a knowledge graph of the ideas that really helped me visualize how concepts tied together.”
After identifying hypotheses, Codex writes the scripts on machine learning infrastructure to train models. The Codex app supports SSH, so Joshi has found that he no longer has to worry about log-in and setup on remote hosts; he can easily run large machine learning workloads from his laptop.
“If you have an old codebase that isn’t that performant, Codex is really good at machine translation. So a lot of folks are taking their Python repository, sending it to GPT-5.5, and it’s rewriting it into Rust and making it like 20X more efficient.”
—Shaunak Joshi, AI researcher
What’s next
Codex is accelerating work at NVIDIA across engineering and research teams, taking ideas from concept to execution and testing in a single workflow.
“We’re just scratching the surface of what it can do,” Hannusch says. “I’m really excited to keep building real systems and see how far it can go.”