FaceID Inventor's AI Startup Aims to Revolutionize Brain Health

Former Apple FaceID and Vision Pro co-inventor Gidi Littwin is making waves in the artificial intelligence sector with his startup, Hemispheric. The company has secured $52 million in funding to advance its frontier AI model designed to analyze the brain's electrical activity, aiming to diagnose neurological and psychiatric conditions such as depression, PTSD, and Parkinson’s.

Littwin, who departed Apple in 2020, teamed up with Hagai Lalazar, who had already begun developing AI for non-invasive brain study. Their shared vision is to transform brain health diagnostics, making it as accessible and straightforward as a routine blood test. This ambitious goal leverages Littwin's extensive experience in building deep-learning models trained on massive datasets during his tenure at Apple.

Pioneering Brain Data Collection

The foundation of Hemispheric’s innovative approach lies in its vast data collection efforts. Similar to the extensive data gathering Littwin oversaw for FaceID and Vision Pro, Hemispheric amassed a quarter-million hours of brain data from 100,000 paid volunteers across Asia, Tel Aviv, and Boston. These participants engaged in game-like activities specifically designed to activate various brain regions.

This rich dataset was crucial for training Hemispheric’s frontier model. The AI system learns to infer brain function from electrical signals within the skull, mirroring how large language models deduce meaning from analyzing text statistically. This breakthrough allows for an objective, data-driven assessment of brain health, moving beyond traditional subjective questionnaires and behavioral observations currently used for many cognitive disorders.

Transforming Diagnostics and Treatment

Hemispheric has already demonstrated the model's efficacy by testing it on subsets of individuals diagnosed with conditions like PTSD, schizophrenia, and depression, reporting accurate deductions about their brain health. The team is also conducting a clinical study to evaluate the model’s potential to diagnose and even predict Alzheimer’s disease.

The company plans to submit its first product, aimed at studying PTSD, to the FDA for approval early next year. If successful, public rollout is anticipated later in 2027. The diagnostic process involves a patient wearing a lightweight EEG headset for about 15 minutes, interacting with a tablet application. The AI model then helps clinicians interpret these electrical signals to make diagnoses, predict treatment effectiveness, and monitor patient progress.

Funding and Future Ambitions

The $52 million in early-stage funding, sourced from American and Israeli venture capital firms and individual investors like early Uber-backer Howard Morgan, will fuel Hemispheric’s expansion. The capital will be used to forge partnerships with governments, healthcare organizations, and pharmaceutical companies, expand hiring in the US, and accelerate regulatory approvals.

Looking ahead, Hemispheric intends to collect even more brain data from millions of people to continually refine its model. Furthermore, the company is actively developing its own brain scanners. Littwin notes that traditional EEGs were not designed for machine learning, and these proprietary devices aim to provide more useful data specifically optimized for deep learning applications, promising even greater diagnostic precision in the future.

FAQ

Q: What is Hemispheric's primary goal with its AI technology? A: Hemispheric aims to revolutionize brain health diagnostics by using a frontier AI model to analyze brain electrical activity, enabling the non-invasive diagnosis and monitoring of cognitive disorders such as depression, PTSD, Parkinson's, and potentially Alzheimer's. The company envisions this technology being as simple and affordable as a blood test.

Q: How does Hemispheric's AI model acquire and interpret brain data? A: The company collected a vast dataset of brain activity—a quarter-million hours from 100,000 volunteers—who performed activities designed to stimulate different brain regions. This data trained an AI model that infers brain function from electrical signals, similar to how large language models interpret text. Patients would use a lightweight EEG headset and a tablet app, with the AI assisting clinicians in decoding the signals.

Q: When can we expect Hemispheric's diagnostic product to be available to the public? A: Hemispheric plans to submit its first product, focused on PTSD, for FDA approval early next year. If regulatory processes proceed successfully, the company anticipates rolling out the technology to the public later in 2027.