Apple FaceID co-inventor raises $52M for brain-decoding AI model
What's the deal? Hemispheric, a startup co-founded by Apple FaceID and Vision Pro co-inventor Gidi Littwin, has raised $52 million. The company has spent six years building a frontier AI model that decodes electrical activity in the brain to help diagnose cognitive disorders.
What's the endgame? Hemispheric wants to examine the brain without invasive procedures. Its model infers brain function from electrical signals inside the skull — much as large language models deduce meaning by statistically analysing text.
Why now? Because each brain looks different, doctors have largely relied on subjective questionnaires and behavioural observations to diagnose depression, Alzheimer's, and Parkinson's. Hemispheric aims to replace that guesswork with data.
To train the model, Littwin and cofounder Hagai Lalazar collected what they call their “most prized possession”: a quarter of a million hours of brain data from 100,000 paid volunteers across Asia, Tel Aviv, and Boston. Subjects completed game-like activities that engaged different parts of the brain.
The founders met after Lalazar, who had begun developing AI to study the brain without surgery, cold-messaged Littwin on LinkedIn. Lalazar had spoken to around 75 candidates before finding a commercially minded partner. Littwin, who left Apple in 2020, had experience with the “massive data collection operations” behind FaceID and Vision Pro hand-tracking.
“We knew we had to build something very similar at Hemispheric,” Littwin said, “and we have.”
What could go wrong? The company tested its generalised model on people diagnosed with PTSD, schizophrenia, and depression, and said it made accurate deductions about their brain health. But those are early results. Hemispheric is now running a clinical study to validate the technology — a step that could determine whether the model holds up under scrutiny.
The signal: Hemispheric is applying the data-and-scale playbook of large language models to neuroscience, betting that a foundation model trained on enough brain data can turn subjective diagnosis into measurement. If it works, the approach could reshape how cognitive disorders are identified.
Read more: RedHot
Image credit: jurvetson
