Code Puppy Creator Leaves Walmart for AI Startup Pydantic

Michael Pfaffenberger, a distinguished engineer at Walmart Global Tech, has announced his departure from the retail giant. Over the course of a year, he built and grew Code Puppy, an internal tool for writing code through natural-language commands that is now used by nearly 75,000 Walmart employees. Pfaffenberger is joining Pydantic, the startup behind the technology his tool is built on.

From internal experiment to company-wide tool

Code Puppy started as an internal project meant to simplify writing, editing and testing code through simple text commands. Over twelve months, the tool went from a thousand users to nearly 75,000 Walmart employees using it daily. The company also released an open source version of the project, which has been downloaded more than 496,000 times, with a record month reaching 71,000 downloads.

The scale of interest goes well beyond the IT department. As Walmart Chief Technology Officer Suresh Kumar noted, the tool is used by merchandising and supply chain staff as well as store managers, people who previously had no direct exposure to writing code. That effect, the democratization of coding inside one of the world's largest companies, has made Code Puppy a flagship example of so-called vibe coding in a corporate setting.

Why Pydantic

Code Puppy was built on Pydantic AI, a framework for building AI agents developed by the British startup Pydantic, previously known mainly for its popular Python data-validation library. The company, founded by Samuel Colvin, has evolved from an open source project into a commercial platform that includes the Pydantic AI agent framework, the Logfire observability system and the Pydantic Evals model-evaluation tool.

Pydantic Chief Technology Officer David Montague said Pfaffenberger's experience building AI tools that coordinate multiple agents at once inside a large enterprise is a rare and valuable skill today, one that companies trying to scale agent deployments beyond isolated pilot projects are eager to secure.

I'm filled with joy that I got to build software that improved so many people's lives, at work and beyond - Michael Pfaffenberger, creator of Code Puppy

People use this in merchandising, in the supply chain, store managers use it - Suresh Kumar, Walmart Chief Technology Officer

What happens to Code Puppy next

Walmart confirmed Pfaffenberger's departure and said the project will continue under the team taking over responsibility for the tool. Those named include Ryan Eccles, John Choi, Kaveri Bakshi, Wesley Blakemore, Andrew Tilson and TJ Webb. In a farewell note to colleagues, Pfaffenberger himself stressed that Code Puppy is staying at the company and will keep being developed regardless of his own decision to change jobs.

What it means for the AI coding tools market

The Code Puppy story shows how large corporations are increasingly building their own internal tooling layers on top of open AI frameworks instead of waiting for ready-made commercial products. For startups like Pydantic, the success of such a deployment is also proof that their technology holds up at the scale of thousands of corporate users, not just in demos and pilot projects.

For Polish companies and IT teams considering similar projects, it's a signal that internal vibe-coding tools can meaningfully shrink the gap between non-technical employees and the ability to automate their own tasks. At the same time, Walmart's case highlights the risk tied to such initiatives: a project's success can end up tightly bound to a single person, and that person's departure raises questions about continuity and the pace of future rollouts.

Pfaffenberger's move also fits a broader trend of experienced engineers leaving large corporations for smaller, specialized AI startups, where they can directly shape the development of tools they previously used as internal customers. For Pydantic, bringing in someone who stress-tested its framework in practice across tens of thousands of users amounts to a direct transfer of knowledge about what works and what needs fixing in a production environment.