Sbir phase II: AI-mediated neurotechnology for enhanced access to virtual and physical reality

The National Science Foundation awarded Cooperative Agreement 2545788 to In Virtualis LLC, a SBA-certified women-owned small business, for $1.25 million on August 15, 2026, under the NSF Technology, Innovation, and Partnerships program (CFDA 47.084).

The award funds development of an artificial intelligence middleware layer that converts minimal biosignals—such as brain-computer interface and electromyographic inputs—into context-aware, multi-step actions. Current biosignal systems provide only two to ten bits per second of control information, insufficient for precise task execution; the middleware bridges this intent-to-action gap by augmenting sparse biosignals with task-relevant contextual information. The platform incorporates three core technical innovations: real-time affordance prediction that infers contextually appropriate actions from video and gaze data, hierarchical action decomposition that maps simple user inputs to multi-step behavioral sequences, and adaptive user modeling that personalizes control strategies through reinforcement learning. This enables users to complete full activities of daily living through virtual and physical reality interfaces. The recipient intends to commercialize the solution through runtime licensing to original equipment manufacturers, supporting rapid adoption across existing distribution channels and clinical and research deployments.

Work is performed in Mercer Island, Washington, with a period of performance through July 31, 2028.

Abstract

The broader commercial impact of this sbir phase II project is to develop an artificial intelligence middleware layer that transforms minimal biosignals into context-aware, multi-step actions, enabling users to complete full activities of daily living. This innovation advances scientific understanding of intent-based human computer interaction while addressing a critical GAP in the technology market. The middleware architecture provides a competitive advantage by offering device manufacturers a standardized, interoperable solution that reduces in-house engineering costs and accelerates deployment, allowing firms to concentrate on core hardware innovation. This commercialization strategy leverages runtime licensing to original equipment manufacturers, supporting rapid adoption through existing distribution channels and alignment with responsible AI deployment practices. Initial deployments target specific manufacturers serving research and clinical institutions. This sbir phase II project addresses the fundamental bandwidth mismatch between low-dimensional biosignal inputs and complex control demands. Current brain computer interface and electromyographic systems typically provide only two to ten bits per second of control information, which is insufficient for precise task execution. Building on phase i results, the project advances a modular artificial intelligence middleware architecture that bridges this intent-to-action GAP by augmenting sparse biosignals with task-relevant contextual information. The platform introduces three core technical innovations: real-time affordance prediction that infers contextually appropriate actions from video and gaze data, hierarchical action decomposition that maps simple user inputs to multi-step behavioral sequences, and adaptive user modeling that personalizes control strategies through reinforcement learning. Technical development targets sub five hundred millisecond end-to-end latency, intent prediction accuracy of at least eighty five percent, and seamless interoperability across heterogeneous hardware platforms. The system employs edge computing for latency-sensitive inference with cloud-based resources supporting model updates and continuous improvement. Anticipated outcomes include standardized integration protocols for biosignal acquisition systems, validated performance benchmarks, and live demonstrations of the middleware controlling robotic end effectors. Collectively, this work establishes foundational principles for amplifying sparse biological control signals into functionally complete behaviors, advancing both the scientific understanding and practical deployment of next-generation human machine interfaces. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.