A startup founded by one of the engineers behind Apple's FaceID has raised $52 million after building deep learning models trained on brain data collected from 100,000 people, an approach it believes can finally make brain health measurable without surgery. The company, called Hemispheric, wants its technology to eventually diagnose and monitor conditions such as depression, Alzheimer's and Parkinson's using nothing more invasive than a headset a patient wears for about 15 minutes.
From FaceID and Vision Pro to Brain Data
Gidi Littwin, who helped build FaceID at Apple, left the company in 2020 looking for a new challenge. He found it when Hagai Lalazar, who would become his Hemispheric cofounder, cold-messaged him on LinkedIn. Lalazar had already started developing artificial intelligence that could study the brain without surgery and was searching for a commercially minded partner to help turn the technology into a company; by the time he reached out to Littwin, he had already spoken with around 75 other candidates.
At Apple, Littwin had also worked on hand-tracking for the Vision Pro augmented reality headset, a project that required collecting what he described as hundreds of thousands of subjects' worth of data to train the deep learning models behind the feature. That experience shaped how the two cofounders approached Hemispheric. "There were massive data collection operations behind these projects and we knew we had to build something very similar at Hemispheric," Littwin says, "and we have."
A Quarter-Million Hours of Brain Data
Doctors diagnosing conditions like depression, Alzheimer's and Parkinson's have long had to rely mostly on subjective questionnaires and behavioral observation, in part because every person's brain activity looks different. To try to change that, Littwin and Lalazar built what they call their most prized possession, a quarter of a million hours of brain data gathered from 100,000 paid volunteers across Asia, as well as in Tel Aviv and Boston. Subjects took part in a series of activities designed to look like games, but each one was built to activate a different part of the brain.
Teaching AI to Read Electrical Signals
That dataset was used to train what Hemispheric calls a frontier model, which infers brain function from electrical activity recorded inside the skull, in much the same way large language models deduce meaning by statistically analyzing text. The team then tested the generalized model on smaller subsets of people, including those already diagnosed with PTSD, schizophrenia and depression, and said the model made accurate deductions about their brain health. Hemispheric is now running a clinical study to test whether the same model can diagnose, and even predict, Alzheimer's.
First Target: FDA Clearance for a PTSD Tool
The company plans to submit its first product, aimed at studying PTSD, to the FDA for approval early next year, with hopes of bringing it to the public later in 2027. In practice, diagnosing a cognitive disorder involves a patient wearing a lightweight EEG headset that records electrical activity in the brain for around 15 minutes while they interact with an app on a tablet. Hemispheric says its AI model then helps clinicians decode those signals to make a diagnosis, choose the most effective treatment by predicting how a patient will respond, and track progress over time.
Akin to a Blood Test
Lalazar describes the long-term vision in simple terms. "The future that we envision is one where this is akin to a blood test," he says. "The device is going to be very, very cheap; it will be able to be sold and distributed throughout mental health clinics, hospitals, and even psychologists' offices."
Competing With Big Tech's Health Ambitions
Hemispheric is entering a field that is already moving fast. AI-assisted diagnostic tools for conditions such as lung cancer are already in clinical use and are speeding up access to treatment across Europe. At the same time, AI giants including OpenAI and Anthropic are expanding into health care, adding to the competition facing the growing number of startups working in the space.
The Money, and What Comes Next
Hemispheric's $52 million came from early-stage investors including American and Israeli venture capital firms as well as individual investors, among them Howard Morgan, an early backer of Uber. The company plans to use the funding to build partnerships with governments, healthcare organizations and pharmaceutical firms, hire more staff in the US, and push forward its regulatory approval process. It also intends to collect brain data from millions more people to keep improving its model.
Littwin and Lalazar are separately developing their own brain scanners, which they believe can capture more useful data for their models than traditional EEG equipment. "These devices were never built for machine learning and definitely not deep learning," Littwin says.
