NeuroDx.ai confirmed that it will formally submit its in-house developed 400-million-parameter AI model to the Government of India for comprehensive evaluation and validation, under emerging standards for sovereign artificial intelligence assets. The company named the model as Manas-1, a 400-million-parameter Brain Language Model trained on over 25000 subjects and 60000 hours of EEG data. The model is using EEG as a unified biological language; through which the proprietary AI interprets neural patterns to decode brain language.
Using Manas-1, the company has achieved accuracy of over 95% for identifying biomarkers in Epilepsy and other mental health disorders. The model is part of a broader research and product portfolio aimed at establishing India’s first foundation-scale physiological state model. It is capable of interpreting human biosignals for advanced health and neurological understanding.
Dr Siddharth Panwar, Founder Neurodx.ai (Stanford and IIT trained AI expert) said, “We are thankful to the Government of India for supporting our research and acknowledging our work. Our foundational architecture has been engineered to natively process multimodal human biosignals enabling the model to infer a unified physiological state from raw continuous data. This approach moves beyond isolated signal interpretation by incorporating cross-signal reasoning, signal fusion, and clinically informed feature extraction derived from collaborative work with neurologists and physiologists.
“We aim to create a disruption in the diagnostic industry, especially in neuroscience, which can ultimately help in initial diagnosis, followed by more accurate treatment. Our aim is to reach 2 billion parameters in the Brain Language Model in the next few months,” he further added.