Why Pipeworx exists

My last role at Meta was building a team to figure out how Meta’s own infrastructure worked.

At that scale, nobody understands the entire system. It is too large, changes too quickly, and grew too organically for any one person to hold in their head. So Meta staffed a team to investigate it, document it, and make it understandable to the people who depended on it.

Pipeworx does that for the world’s public data.

The SEC, FDA, Federal Reserve, national statistics offices, procurement portals, patent registries, and thousands of other institutions all publish valuable information. But every source works differently: different formats, terminology, access methods, identifiers, and ideas about what constitutes an error.

Pipeworx turns 1,462 of those sources into a single, consistent connection that AI agents can discover and use.

That matters because a model that cannot find an answer will often produce one anyway—and the invented answer can sound exactly like the real one. Pipeworx returns information from the source, along with where it came from and when it was retrieved. Every answer can be traced, checked, and challenged.

That is the product: live data for AI agents, with evidence.

Background

Before Pipeworx, I worked at Meta, AWS, and Oculus and founded several companies of my own. I have spent twenty-five years in software, much of it making complicated systems understandable and usable—through documentation, developer relations, localization, and the teams that do that work.

Mojibake

Mojibake is the company behind Pipeworx. Pipeworx is its first product and the live-data infrastructure on which we are building several others.

Questions, partnership ideas, or a data source you think Pipeworx should support? Get in touch.