Non-invasive speech brain-computer interfaces
Every thought
deserves a voice
We're building the technology to restore speech for people who have lost it. Non-invasive. Scalable. Built on the principle that more data and better models will always win.
The problem
Speech is the most
human thing we do
Every year, hundreds of thousands of people lose the ability to speak. ALS, stroke, traumatic brain injury, neurodegenerative disease — the causes are varied, but the outcome is the same: isolation from the people and world around you.
Current assistive technologies are slow, frustrating, and fundamentally limited. People who once spoke hundreds of words per minute are reduced to painstakingly selecting letters on a screen. We believe they deserve better.
5M+
people worldwide
living without speech
Our approach
The bitter lesson,
applied
“The biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective.”
Most approaches to neural speech decoding rely on hand-crafted feature engineering: carefully designed signal processing pipelines, domain-specific architectures, and painstaking per-patient calibration. These approaches plateau.
We take a different path. We build large-scale datasets of neural activity paired with speech. We train foundation models that learn the deep structure of how the brain encodes language. And we let scale do what it always does — unlock capabilities that hand-engineering never could.
01
Scale data
We build the largest non-invasive neural-speech datasets ever assembled, across diverse populations and conditions.
02
Scale models
Foundation models trained on massive neural data learn representations that generalize across individuals and tasks.
03
Speech emerges
With enough data and the right architecture, real-time speech decoding doesn’t need to be engineered — it emerges.
Technology
Non-invasive
by design
Our system reads neural signals from the surface of the scalp — no surgery, no implants, no risk. This isn't a compromise. It's a deliberate choice: the technology that will actually reach millions of people must be one that any clinic can deploy and any patient can use.
Zero surgical risk
Surface-level neural recording
Generalizable
Models transfer across individuals
Real-time
Thought to speech in milliseconds
Accessible
Deployable in any clinical setting
Join us
We're looking for people
who think in decades
If you're a researcher, engineer, or clinician who believes the future of neural interfaces is non-invasive and data-driven, we want to talk. We're building a team as ambitious as the problem we're solving.