Ex-Apple Face ID Engineer Raises $52 Million to Build AI That Reads Brain Signals

Hemispheric, the brain-technology startup founded by former Apple engineer Gidi Littwin and cofounder Hagai Lalazar, has raised $52 million to develop an AI system that analyzes brain activity and could help diagnose cognitive and psychiatric disorders without invasive procedures. The company says it has already collected brain data from 100,000 paid volunteers and is preparing its first FDA submission for a PTSD-related product early next year.

The funding round underscores how rapidly artificial intelligence is moving into healthcare, particularly into diagnostics that have traditionally depended on physician observation, patient questionnaires, and slow, expensive testing. Hemispheric’s pitch is ambitious: build a frontier model for the brain, then use it to turn a short EEG session into something closer to a precision clinical test.

Why Hemispheric’s new funding round matters

Hemispheric’s financing gives the company more firepower to pursue one of the most difficult goals in digital health: translating raw brain signals into clinically useful insight at scale. The startup is not just building software for clinicians. It is trying to create the data infrastructure, hardware, and regulatory pathway needed to make brain-reading AI practical in everyday care.

That matters because disorders such as depression, PTSD, Alzheimer’s disease, schizophrenia, and Parkinson’s are still difficult to diagnose cleanly. In many cases, doctors must make decisions based on interviews, symptoms, and observed behavior rather than a clear biological marker. Hemispheric believes it can change that by treating brain activity like a pattern-recognition problem that large neural networks are uniquely suited to solve.

The startup’s backers include American and Israeli venture firms as well as individual investors such as Howard Morgan, an early Uber supporter. The company plans to use the capital to accelerate regulatory work, deepen partnerships with governments and healthcare organizations, and expand hiring, especially in the United States.

Who is Gidi Littwin, and why did he leave Apple?

Gidi Littwin is an engineer best known for his work on Apple’s Face ID system and later on hand tracking for the Vision Pro headset. He left the company in 2020 looking for a different kind of challenge, and he found one when a LinkedIn message from Hagai Lalazar led him into brain AI.

According to the company, Lalazar had already spent months speaking with dozens of potential cofounders before reaching Littwin. What made the pair click was a shared conviction that high-quality, large-scale data collection would be the difference between a research idea and a clinically credible product.

Littwin has said that Apple’s major product efforts depended on enormous data-collection operations, and that Hemispheric would need a similar discipline if it wanted its models to work in the real world.

That experience at Apple appears to have shaped Hemispheric’s operating model. Rather than relying on a small dataset from a narrow clinical population, the company built a huge repository of brain recordings from people across multiple geographies and backgrounds. In Littwin’s view, scale is not optional if the company wants to identify reliable signals amid the variability of human brain activity.

How does Hemispheric’s brain AI work?

Hemispheric’s system is built to infer patterns from electrical activity inside the skull, using machine learning methods that the company compares to how large language models find meaning in text. In practical terms, the model is trained to look for statistical patterns in brain signals and map them to likely neurological or psychological states.

The process starts with a lightweight EEG headset, which records electrical activity for roughly 15 minutes while a person completes interactive tasks on a tablet. Those tasks are designed to feel like games while activating different regions and circuits in the brain. The output is then analyzed by Hemispheric’s AI system, which the company says can help clinicians make diagnoses, choose interventions, predict treatment response, and monitor progress over time.

What makes the model different from traditional diagnosis?

Hemispheric’s model is intended to reduce reliance on subjective judgment. For many cognitive and psychiatric conditions, doctors still depend heavily on questionnaires and behavioral observation because the brain is difficult to measure directly in a quick, noninvasive way. The startup believes its model can turn a short recording session into a more objective diagnostic signal.

The company says it has already tested the generalized model on groups of people with PTSD, schizophrenia, and depression and found that it produced accurate assessments of brain health. It is now running a clinical study aimed at determining whether the approach can also identify Alzheimer’s disease and potentially forecast it before symptoms become more obvious.

That last point could be particularly important. If successful, a system that flags risk earlier could help clinicians intervene sooner, select more appropriate treatments, and track disease progression with far more granularity than is typical today.

What data did Hemispheric use to train its model?

Hemispheric says its model was trained on what the founders call the company’s most valuable asset: a quarter of a million hours of brain data collected from 100,000 paid volunteers. The data came from participants in Asia, Tel Aviv, and Boston, giving the company a geographically and demographically broad dataset by the standards of brain research.

The volunteers completed a series of structured activities that were designed to stimulate different parts of the brain. The company argues that the breadth of this data matters because brain signals vary dramatically from one person to another. A model trained on a narrow dataset would struggle to generalize across populations or clinical settings.

That scale also helps explain why the company has spent years in development before moving toward a commercial launch. Collecting enough data to train a model for the brain is not a quick software project. It requires recruiting participants, standardizing protocols, validating the quality of recordings, and building an analysis pipeline robust enough for clinical use.

How close is the company to a real medical product?

Hemispheric is already moving beyond research toward regulation. Its first product, focused on PTSD, is expected to be submitted to the FDA early next year. If cleared, the company hopes to make the product publicly available later in 2027.

That timeline suggests Hemispheric is still in the difficult middle stage that many health AI companies face: promising enough to attract major capital, but not yet validated through broad clinical deployment. The FDA process will be a key test of whether the company’s data and model performance are strong enough to support medical claims.

For now, the product’s use case is relatively simple to describe. A patient wears the headset, completes the tablet-based session, and the software helps clinicians interpret the recording. Over time, the company wants the system to do more than categorize symptoms. It wants to contribute to treatment selection and longitudinal monitoring, which would make it more valuable in routine care.

Why is Hemispheric building its own brain scanners?

Hemispheric is also developing its own scanners because it believes conventional EEG devices were not designed for modern machine learning, let alone deep learning. The founders argue that the hardware itself may be limiting the quality and utility of the data.

Littwin has said that traditional devices were created for older clinical workflows rather than the kind of data-rich model training Hemispheric needs. That is a familiar theme in AI: if the sensor is noisy, incomplete, or poorly optimized for the task, the model can only do so much with it.

By controlling both the hardware and the software, the company hopes to improve signal quality and generate more informative recordings. That would potentially make the AI better at spotting patterns that ordinary EEG systems miss.

What does Lalazar mean by a “blood test” for the brain?

Lalazar is describing a future in which brain diagnostics become as routine, cheap, and widely available as common bloodwork. In that vision, a clinician would not need specialized imaging or a complex hospital procedure to get a useful readout. Instead, a brief session with lightweight equipment and AI analysis could provide actionable information.

The comparison is aspirational, but it captures the commercial logic behind Hemispheric’s strategy. If the product can be made inexpensive, fast, and portable, the company could sell it to mental health clinics, hospitals, and private practices rather than limiting it to major medical centers.

Lalazar has said the device should eventually be low-cost and easy to distribute across mental health clinics, hospitals, and psychologists’ offices.

How competitive is the market for AI in healthcare?

The market is increasingly crowded, and Hemispheric is arriving at a moment when AI is becoming mainstream in clinical workflows. Systems that assist with lung cancer diagnosis are already being used in parts of Europe, where they are helping speed access to treatment. At the same time, major AI companies are pushing deeper into healthcare.

OpenAI and Anthropic are both expanding into the sector, adding pressure on startups that want to build specialized medical tools. Their presence matters because big-model companies bring technical talent, strong brand recognition, and access to enterprise customers. Smaller firms like Hemispheric must therefore prove that their niche expertise and proprietary datasets can give them an edge.

Hemispheric’s advantage, at least on paper, is focus. Rather than building a general healthcare chatbot or broad diagnostic assistant, the company is centered on a narrow and technically demanding domain: the brain. That specialization could make its model more useful in a specific clinical workflow, even if the market is smaller than general-purpose healthcare AI.

Why brain AI is so difficult

Brain AI is hard because the organ itself is complicated, noisy, and highly individual. Two patients with similar symptoms may produce very different signals, and the same person’s patterns can change depending on medication, sleep, stress, or the way a test is administered. That variability makes it difficult to build a model that is both accurate and generalizable.

There is also a clinical challenge. A diagnosis is only as useful as the action it enables. For a brain AI system to matter in practice, it must not only detect patterns but help doctors decide what to do next. That means evidence from well-designed clinical studies, not just promising internal testing.

Hemispheric appears to understand those hurdles. Its strategy combines large-scale data gathering, model training, hardware development, and regulatory planning. In other words, the company is trying to solve the full stack problem rather than just the algorithmic one.

What risks does Hemispheric still face?

The biggest risks are clinical validation, regulatory approval, and adoption by physicians. The company may have an impressive dataset, but medical tools must show they are reliable across real-world populations, not just in controlled testing environments.

Another risk is trust. Mental health and neurological diagnosis touch sensitive, often life-changing decisions. Clinicians will want clarity on accuracy, bias, false positives, false negatives, and how the model performs in different age groups, geographies, and disease stages.

There is also the question of workflow. Even a strong product can fail if it is cumbersome to use, hard to reimburse, or too expensive for routine care. Hemispheric’s promise of a lightweight, cheap, fast test is aimed directly at that concern, but the company still has to prove the promise in practice.

What the funding means for the next phase

The $52 million raise should give Hemispheric enough runway to keep building through the regulatory and clinical trial stages. The company says it will use the money to expand partnerships with public institutions and healthcare providers, strengthen its US presence, and continue gathering data at a larger scale.

That last point is important because the founders do not appear to see 100,000 volunteers as the endpoint. They want to collect brain data from millions of people eventually, a scale they believe will improve the model’s performance and broaden its applicability. In AI terms, they are betting that more data will continue to sharpen accuracy and robustness.

If the startup is successful, it could emerge as a significant player in a new category of neurologic diagnostics. If it is not, it may still influence how the sector thinks about the combination of high-volume data collection, wearable hardware, and deep learning in medicine.

Timeline: from Apple engineering to FDA review

Hemispheric’s evolution shows how quickly a consumer-tech background can be redirected toward life sciences when paired with capital and a clear scientific hypothesis. The following timeline captures the main stages so far.

• Apple years: Littwin contributes to Face ID and later Vision Pro hand tracking.
• 2020: He leaves Apple seeking a new direction.
• Founding period: Lalazar recruits him after extensive outreach to potential partners.
• Data collection phase: The company gathers 250,000 hours of recordings from 100,000 volunteers.
• Model development: Hemispheric trains a generalized brain AI system on those recordings.
• Clinical testing: The company evaluates performance in groups with PTSD, schizophrenia, and depression.
• Near-term milestone: FDA submission planned for a PTSD product early next year.
• Commercial target: Public rollout could follow in 2027 if approval proceeds.

What happens next?

The next 18 months will likely determine whether Hemispheric becomes a serious medical company or remains a highly promising research story. The FDA submission will be the first major external checkpoint, but the company also needs strong study results, durable partnerships, and a clear path to adoption.

Still, the size of the funding round and the ambition of the platform suggest investors are willing to back the idea that brain diagnostics can be improved by the same kind of data-intensive AI that transformed language, vision, and speech. Hemispheric is now trying to prove that the same logic can be applied inside the skull.

If it works, the result could be a cheaper, faster, more objective way to assess cognitive and psychiatric conditions. If it does not, the effort may still help define the next generation of brain-tech tools. Either way, the company’s progress will be watched closely as AI pushes deeper into medicine.