What You Should Know
- NeuroAI foundation company Hemispheric has officially emerged from stealth, announcing a $52M early-stage funding round co-led by Hanaco Ventures and OneMind/Awareness Capital to scale objective human brain decoding.
- The company has launched Descartes, a 6-billion-parameter neural foundation model that translates non-invasive, dry-electrode EEG electrical waves into quantitative, actionable clinical insights.
- Disrupting traditional sample-size limitations, Descartes was trained on a proprietary multi-modal dataset spanning 250,000 hours of continuous EEG and behavioral metrics collected from over 100,000 active participants.
- Co-founded by the co-inventor of Apple’s FaceID, Hemispheric’s clinical triage layer requires no surgery—relying entirely on a 15-minute, non-invasive headset check paired with automated mobile app micro-tasks.
- The platform’s initial go-to-market applications deliver instant structural profiles to optimize precision medicine pathways for severe, hard-to-track central nervous system (CNS) conditions including PTSD, mTBI, depression, and Alzheimer’s disease.
Cracking the Neural Scaling Law: The Architecture of Descartes
The technology framework underlying Descartes moves completely past simple signal-processing scripts to prove that brain decoding follows the exact same computational scaling laws that unlocked large language models and computer vision. Co-founded by computational neuroscientist Hagai Lalazar, PhD, and Gidi Littwin—the elite deep learning architect who co-founded RealFace and served as the co-inventor of Apple’s FaceID infrastructure—Hemispheric built its system of record completely from scratch.
The company spent six years constructing proprietary, clinical-grade data collection laboratories and highly specialized computing environments explicitly optimized for dense neural data processing.
To eliminate the deep signal variability that historically compromised non-invasive electroencephalography (EEG)—where identical neural messages display radically different waves across different human skulls—Hemispheric compiled an unprecedented dataset.
The 6-billion-parameter Descartes model was trained on more than 250,000 hours of multimodal EEG and behavioral recordings extracted from over 100,000 unique participants.
By analyzing cognitive, emotional, and perceptual brain activity at this massive scale, Descartes achieves zero-shot generalization. It effectively translates the brain’s raw electrical currents into an objective, quantitative profile of real-world neural function, isolating hidden disease markers long before outward physiological symptoms manifest.
Frictionless Diagnostics: Transitioning Neurotech from Labs to Sidelines
By completely rejecting high-cost, invasive neurosurgical procedures, lengthy hospital testing environments, and surgical implants, Hemispheric delivers an unassailable go-to-market advantage that requires no specialized infrastructure to deploy.
Instead, a patient simply dons a lightweight, dry-electrode EEG headset for 15 minutes while interacting with targeted application tasks on a standard tablet or smartphone.
The custom sensor network captures full-head electrical signals under operational load, enabling Descartes to map distinct psychiatric and cognitive phenotypes in minutes. This frictionless method allows primary care offices, local sports groups, and decentralized clinical trials to rapidly capture clear neural baselines, distinguish between complex disorder subtypes (such as isolating specific depression sub-categories), and closely track the efficacy of targeted treatment interventions over time.
Having already presented the platform’s architectural data to the leadership of the FDA’s Center for Devices and Radiological Health (CDRH), Hemispheric is aggressively executing a broad regulatory strategy to establish the first objective vital sign for precision brain health.
