The engineer behind Apple Face ID now wants AI to read your brain

Key points

• Hemispheric, a neurotech startup, raised $52 million in early-stage funding as of 2025.
• The company collected 250,000 hours of brain-activity data from 100,000 paid volunteers across Asia, Tel Aviv, and Boston to train its AI model.
• A patient wears a lightweight EEG headset, a sensor cap that reads electrical signals from the scalp, for roughly 15 minutes while completing tablet-based tasks.
• Hemispheric plans to submit its first product, focused on PTSD, to the US Food and Drug Administration for approval in early 2026, with a public launch targeted for late 2027.
• The company also has an Alzheimer's detection and prediction study underway.

Gidi Littwin helped build two things you probably use without thinking: Face ID, the facial-recognition system that unlocks iPhones, and the hand-tracking software inside Apple's Vision Pro headset. Both required collecting data from hundreds of thousands of people to teach AI what human faces and hands look like. He left Apple in 2020, and now he is applying the same logic to something far more personal: the human brain.

His startup, Hemispheric, wants to give doctors a reliable, non-invasive way to diagnose conditions like depression, Alzheimer's disease, and PTSD, post-traumatic stress disorder, a condition triggered by severe psychological shock. Today, those diagnoses rely heavily on questionnaires and a clinician's observation, because each person's brain activity looks different enough that there is no standard scan to read.

Hemispheric's answer is a two-part system. A patient wears a lightweight EEG headset, short for electroencephalogram, a cap of sensors that picks up electrical signals from inside the skull without cutting the skin. For about 15 minutes, the patient interacts with tasks on a tablet, tasks that look like simple games but are designed to activate specific brain regions. The AI model then analyses the electrical patterns and helps a clinician make sense of them.

The AI works similarly to the large language models behind chatbots like ChatGPT: instead of finding patterns in words, it finds patterns in brainwaves. To build it, Littwin and his co-founder Hagai Lalazar, a neuroscientist who cold-messaged Littwin on LinkedIn after approaching roughly 75 other candidates, gathered a quarter of a million hours of brain data from 100,000 paid volunteers. That dataset, which Wired AI first reported on, is what the founders call their "most prized possession."

The team says internal tests on people diagnosed with PTSD, schizophrenia, and depression showed the model made accurate assessments of brain health. The word "says" matters here: those results have not yet been published in a peer-reviewed journal, meaning independent scientists have not scrutinised the methodology.

What does this mean for patients?

For now, nothing changes at the clinic. Hemispheric is still years from your doctor's office. The company plans to file for FDA clearance, the approval US regulators require before a medical device can be sold, for its PTSD tool in early 2026. If that goes smoothly, a public launch could follow in late 2027.

Longer term, co-founder Lalazar describes the goal as something "akin to a blood test": a cheap, widely available device that mental-health clinics and even individual psychologists could buy. The $52 million will fund that regulatory push, US hiring, and partnerships with hospitals and pharmaceutical companies.

The founders are also building their own brain scanners. Standard EEG hardware, they argue, was designed decades before machine learning existed and may not capture the detail their models need.

AI diagnostic tools for conditions like lung cancer are already in clinical use in Europe. Hemispheric is betting the same shift is coming for brain health, and that a former Apple engineer's talent for large-scale data collection might be the edge that gets them there first.