[ About ]
Machine Poetry is an independent AI lab of 42 people across Berlin, San Francisco and Tokyo. We build language models for products that cannot afford to sound like a machine.
[01] — Story
Four years, five chapters, one obsession.
2022
Ines and Theo quit their research jobs to answer one question: can a model have taste?
2023
A 3B-parameter model that wrote better support replies than the humans who trained it. $12M seed.
2024
One thousand teams in the first month. We hire our first poet-in-residence.
2025
$140M to build custom inference silicon and our third hub.
2026
One million tokens, 94 languages, 38 ms. Still one very long kitchen table.
[02] — People
The humans behind the machine.
Co-founder & Chief Scientist
Head of Alignment
Design Director
Head of Language Quality
Research Engineer
Principal Researcher
Staff Engineer, Inference
Co-founder & CEO
01
If a sentence needs a second read, we rewrite the model — not the user.
02
Rhythm is a feature. We tune how answers feel, not only what they say.
03
Every word ships into someone’s real day. We measure what happens next.