Core Technology: State Space Models and the LMU

The foundation of ABR?s technology is the Legendre Memory Unit (LMU), a patented neural network architecture based on the principles of how the brain represents and processes time. * Superior Efficiency: Unlike traditional Recurrent Neural Networks (RNNs) or Transformers, the LMU can process extremely long sequences of time-series data using significantly less power and memory. * Mathematical Precision: The architecture is based on a "State Space" mathematical framework, allowing it to compress and store temporal information with near-perfect accuracy while maintaining a tiny computational footprint.

Products and Offerings

1. Hardware: The Edge Time Series Processor (TSP1)

Launched in late 2025, the TSP1 is ABR's flagship AI chip. It is a tiny, brain-inspired processor designed specifically for battery-powered devices. * Ultra-Low Power: The chip consumes less than 35mW while performing complex tasks like Automatic Speech Recognition (ASR) or Text-to-Speech (TTS). * On-Chip Intelligence: It supports up to 10 million 8-bit parameters, allowing it to store and run full AI models entirely on-device, ensuring user privacy and zero-latency response. * Direct Sensor Connectivity: The chip connects directly to audio and biosensing sensors, reducing the need for external components and simplifying the design for device manufacturers.

2. Software: The Nengo Ecosystem

Before becoming a hardware company, ABR developed Nengo, one of the world's most sophisticated neural modeling and simulation platforms. * Nengo Brain Maker: A Python-based package used to build, test, and deploy large-scale brain models and spiking neural networks. * NengoDL: A deep learning library that allows developers to integrate TensorFlow and Keras models with neuromorphic hardware. * NengoLoihi: A specialized compiler used to run AI models on Intel's "Loihi" neuromorphic research chips.

3. Integrated Solutions and Applications

ABR provides turnkey AI models optimized for its TSP1 hardware and other edge processors: * Voice and Speech AI: Real-time on-device transcription, intent detection, and natural language understanding for AR/VR headsets and smart home appliances. * Biosignal Classification: Advanced health monitoring for wearables, including real-time analysis of ECG, EEG, and other vital signs for fitness and medical diagnostics. * Robotics and Drones: Fast, responsive control systems that allow robots to process sensory data and react instantly in connectivity-constrained environments.

Services and Partnerships

• Custom Model Optimization: ABR works with enterprise clients to port existing AI models to State Space architectures, drastically reducing the power requirements for mass-production hardware.
• Evaluation Kits: The company provides TSP1 evaluation boards and developer documentation for hardware engineers to prototype and test new "Always-On" AI products.
• Venture Capital and Growth: In January 2026, ABR closed an oversubscribed seed funding round led by Two Small Fish Ventures, aimed at scaling the production of its TSP silicon to meet global demand from the consumer electronics and industrial IoT sectors.

Strategic Impact

Applied Brain Research is a key architect of the "Intelligence at the Edge" movement. By enabling devices to "think" locally without relying on energy-intensive data centers, ABR is making AI more sustainable, private, and responsive. Their work represents a major shift from the "Cloud AI" era to an era of pervasive, low-power intelligence embedded in everyday objects.