Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Co-inventor L Apple’s FaceID and Vision Pro technology has spent the past six years building a groundbreaking AI model that could one day help decode electrical activity in the brain to diagnose cognitive disorders.
Now, Geddy Litwin’s startup Hemispheric has raised $52 million in funding after collecting data on the brains of 100,000 people to train deep learning models to scan the brain without the need for invasive procedures.
Littwin left Apple in 2020 looking for a change. He found out when one of his founders in the Western Hemisphere, Hagai Lalazar, sent him a message on LinkedIn. Lalazar set out to develop artificial intelligence to study the brain without the need for surgery, and was looking for a business-minded co-founder to move the company forward. By the time he found Litwin, he had spoken to about 75 candidates.
Littwin helped develop FaceID, and at the time was working on hand tracking for the augmented reality product, Vision Pro. As part of this, he had to collect what he told WIRED was “hundreds of thousands of people’s worth of data” to train the deep learning models that power the technology.
“There was massive data collection behind these projects, and we knew we had to build something very similar at Hemisverek, and we did,” Litwin says.
Because each individual’s brain activity looks different, doctors have had to rely largely on personality questionnaires and behavioral observations to diagnose depression, Alzheimer’s, and Parkinson’s disease. To overcome this, Litwin and Hagai amassed their “most valuable possessions”: a quarter of a million hours of brain data from 100,000 paid volunteers across Asia, as well as Tel Aviv and Boston. Participants did a series of activities that looked like games but activated different parts of their brains.
This data helped train the parametric model, which infers brain function from electrical activity inside the skull in the same way that large language models infer meaning by statistical analysis of text. They then tested the generalized model on subgroups of people, including those diagnosed with PTSD, schizophrenia, and depression, and said the model made accurate inferences about individuals’ brain health. The team is currently working on a clinical study to test whether their model can diagnose and even predict Alzheimer’s disease.
The team will submit their first product, which will be used to study PTSD, to the US Food and Drug Administration for approval early next year. They hope this will allow them to release the product to the public later in 2027.
To help diagnose a cognitive disorder, a patient wears a lightweight EEG headset that measures electrical activity in the brain for about 15 minutes while interacting with an app on a tablet. Hemisverick says her AI model will then help doctors decode the signals to make a diagnosis, choose the most effective intervention by making predictions about treatment, and monitor progress.
“The future we envision is one where it’s more like a blood test,” Lalazar told WIRED in an interview. “The device will be very, very cheap and will be able to be sold and distributed throughout mental health clinics, hospitals and even psychiatrists’ offices.”
AI-powered diagnostic tools for conditions such as lung cancer are already in clinical use and accelerating access to treatment across Europe. Meanwhile, AI giants, including OpenAI and Anthropic, are expanding into healthcare, increasing competition for a wealth of startups in the field.
Hemispheric has raised early-stage funding from investors including US and Israeli venture capital firms and individual investors, including Howard Morgan, one of Uber’s early backers. They will use the funds to strengthen partnerships with governments, healthcare organizations and pharmaceutical companies, hire more in the United States, and work toward regulatory approval. They also plan to measure more brain data from millions of people in an attempt to improve their model
The pair are also developing their own brain scanners to capture information that the company believes can provide more useful data for its models than traditional EEG devices. “These devices were never designed for machine learning, and certainly not for deep learning,” Litwin says.
