To accelerate the software development life cycle, Palo Alto Networks takes advantage of Claude 3.5 Sonnet and Claude 3.5 Haiku on Vertex AI. Claude 3.5 Haiku is a particularly strong fit for code completion, assisting developers by providing auto-complete suggestions as they type, essentially offering the next lines of code. 

Claude 3.5 Sonnet is used for pair programming. For example, if Palo Alto Networks developers ask Claude to “help me improve the performance of this code,” Claude gathers relevant code files and sends them to Claude 3.5 Sonnet, which then returns a code snippet as an answer.

This approach is especially helpful for junior programmers. It can take a new employee up to six months to become fully proficient in the company's extensive and complex codebases. This process also requires a significant investment of time from senior developers, who need to guide newcomers through the code structure.

With Claude on Vertex AI, this onboarding process is much more efficient. Claude understands the overall codebase and is great at code explanation, allowing new hires to get up to speed much faster. They can start contributing code within weeks instead of months, significantly shortening the onboarding time and enhancing overall productivity.

More senior developers use Claude as a pair programmer when they encounter tricky debugging events or need help brainstorming complex architectural solutions. They might say things like, “Hey, check my logic on this,” or “I’m stuck on this bug; can you help me brainstorm?”

So far, the results have been impressive. Patel initially organized a pilot program that included 150 developers from junior to senior roles. “We saw a 20% to 30% increase in feature development and code implementation,” he says.

In addition, the speed of writing unit tests increased by about 10% to 30%. This improvement contributes not only to productivity but also to the overall quality of the codebase, as more comprehensive testing leads to fewer bugs and better software performance.

Google Cloud has also introduced more granular and usage-based Generative AI Scale Units (GSUs). With a minimum commitment period and options for six-month or one-year commitments, Palo Alto Networks can more easily manage costs in the long term.

Another major advantage was seamless SDK integration for cloud provisioned throughput and on-demand services. “We didn’t think that switching from on-demand to provisioned throughput would work easily. But it really did work — flawlessly. All we had to do was change the endpoint, and everything ran smoothly. That was a huge win for us,” says Patel.

Following this success, Palo Alto Networks has rolled out Claude on Vertex AI to 3,000 developers. But Patel and this team aren’t resting on their laurels. They are already looking for more opportunities to boost productivity and security by mapping out the existing software development processes for each developer team.

Another project focuses on implementing AI for post-processing code. In this scenario, a developer writes code using Claude in real-time, and the AI takes over in an offline mode once they complete their work and create a pull request.

Additionally, the AI will enhance code quality by improving variable names, adding comments for clarity, and automatically generating and running unit tests. Then, it identifies security issues with the code and automatically suggests patches to resolve them. “These efforts will contribute to cleaner, more readable code while simultaneously addressing vulnerabilities and strengthening security,” says Patel.

By harnessing the power of Anthropic's Claude models on Google Cloud's Vertex AI, Palo Alto Networks is reimagining its approach to software development. Palo Alto Networks is empowering developers to work more efficiently and effectively, and ultimately accelerating the delivery of secure, high-quality products to customers.