Who's Jetty?

Jetty started with ML engineers who evaluated AI systems for a living, checking whether a run's output actually held up, not just looked fine on the first pass. They kept hitting the same wall: no way to compare one run to the next, and configs that never lived anywhere near the results they produced. Nobody had built the tool for that, so they built it.

"Until now, solving hard problems with AI has meant assembling a whole team," says Angela, Head of Engineering. "I joined Jetty to build that expertise into the platform itself."

That's basically my job description too. My pouch holds a lot more than fish, mostly water, and I only keep what's worth holding onto. Jetty does the same thing with a run: recording it is easy, knowing whether it was any good is the job, and that has to happen the moment the run finishes, by whoever actually understands the work, in a form you can diff and version like any other file.

Jon, founder and CEO, calls Jetty "a one-stop shop for running agents while controlling for cost and quality". These days, the best model for a job keeps changing, so Jetty lets you bring your own agent and model while giving you the advice on when to change it for something better.

The team is small (but mighty!) enough that if you're using Jetty, you're probably talking to someone who helped build it.

Funny enough, I spend my days off on actual jetties, watching the tide and deciding what's worth keeping. Guess the name fit before the company did.