New Way Now: How CodeRabbit uses Google and NVIDIA to automate AI code reviews at scale

*Featured in this video:* David Loker, VP of AI at CodeRabbit and Howard Wright, VP of Startup Ecosystem at NVIDIA

*Executive summary:* As AI-driven coding surges, the volume of pull requests has increased by 30%, overwhelming human reviewers and increasing the risk of bugs in production environments. CodeRabbit works closely with Google Cloud and NVIDIA to build an automated, agentic quality control layer to empower its customers’ development teams. By utilizing Cloud Run for secure, ephemeral sandboxes and NVIDIA Nemotron 3 Super for context gathering along with other frontier models for deep-reasoning, CodeRabbit is reducing the manual burden on developers. The solution enables companies to scale their software production at agentic speeds without compromising on security or code integrity.

*Challenge:* In the era of agentic coding, software development is experiencing a massive shift in velocity. While AI agents can generate syntax at an unprecedented rate, this surge has led to a 30% increase in pull requests (PRs). This growth is remarkable, but it poses a new problem: people can’t review code at the same speed AI is producing it. Without a robust validation layer and review process, companies face a significant risk of bugs and security vulnerabilities slipping into production. CodeRabbit is building that quality control layer to help companies maintain high standards and produce quickly while remaining cost-effective. To do this, the CodeRabbit team needed technology partners that could help them scale.

*Solution:* CodeRabbit built a leading AI code review platform powered by an agentic research pipeline on Google Cloud. By leveraging Cloud Run, it created a system that automatically spawns ephemeral sandbox environments to download and analyze code the moment a PR is opened. This infrastructure allows them to handle burst traffic securely and scale effortlessly using Cloud Tasks. To achieve deep context, 70-80% of the company’s AI work happens before the review even begins—extracting information and exploring the codebase to ensure the AI understands the "why" behind a change. Through the NVIDIA Inception program, CodeRabbit works closely with NVIDIA researchers to optimize these workloads. By utilizing NVIDIA Nemotron 3 Super, they’ve implemented agentic loops that gather the relevant context from dozens of data sources that can be useful in finding bugs in the code being reviewed, significantly raising the bar for review accuracy.

*Result:* By collaborating with Google Cloud and NVIDIA, CodeRabbit has transformed into a high-precision quality control layer for the modern dev stack. By automating the deep research phase of code review, CodeRabbit has dramatically improved the quality of reviews while keeping costs predictable. The team continues to find ways to iterate and improve its recommendations as well. Using NVIDIA-optimized agentic loops has reduced the cost associated with context gathering by well over 50%. Furthermore, by using BigQuery to analyze event streams, CodeRabbit provides customers with total visibility into their development ecosystem. This infrastructure allows CodeRabbit’s customers to embrace the speed of AI coding with the confidence that their code is validated correctly to ensure the AI revolution isn’t slowed down by bugs.

*Highlights and key takeaways from our interview with David Loker, VP of AI at CodeRabbit and Howard Wright, VP of Startup Ecosystem at NVIDIA:*

→ “Google Cloud plays a key role in helping us move faster, so we can focus our attention on the quality of the code review.” — David Loker, VP of AI at CodeRabbit

→ “The processing power needed to mechanize CodeRabbit’s solutions means they’re using our existing tools and then layering on their own unique defensible IP to create something that benefits the world.” — Howard Wright, VP of Startup Ecosystem at NVIDIA

→ “When you're trying to build something brand new, something at the cutting edge of AI, you want to work with the people who know it best. I don’t know better people than the folks at Google and the folks at NVIDIA to build with.”— David Loker, VP of AI at CodeRabbit

*Google Cloud products used:* BigQuery, Cloud Run, Cloud Tasks, Gemini

*Learn more:*
→ 50% faster merge and 50% fewer bugs: How CodeRabbit built its AI code review agent with Google Cloud Run
https://cloud.google.com/blog/products/ai-machine-learning/how-coderabbit-built-its-ai-code-review-agent-with-google-cloud-run?e=48754805