An independent lab that makes AI cheaper and faster to run.
Every answer a model gives costs computation, memory, and time. We work on both sides of that cost: deep learning architectures designed to be efficient from the start, and the engineering that runs them, from GPU kernels to serving, at lower latency. Some ideas remain research. Others become tools or products.
The name
A tile is a simple part of a larger structure. What matters is how the parts fit together and how the system behaves as a whole.
Our mark uses the same visual language: a modular grid, a deliberate cut, and one restrained red line.
How we work
Measure first
We test assumptions and measure behavior before deciding what to optimize.
Keep complexity deliberate
Add dependencies and services when their value outweighs their cost.
Be precise about claims
Distinguish open questions, prototypes, measured results, and available products.
Design for failure
Retries, timeouts, and partial failures are part of the system, not afterthoughts.