Mentium Technologies specializes in developing ultra-low-power, radiation-tolerant AI co-processors designed for size, weight, and power-constrained devices operating in demanding environments such as satellites, spacecraft, drones, robotics, and advanced security systems. Their proprietary architecture enables real-time AI inference with power consumption below 0.5 W, allowing seamless integration into existing hardware without the need for redesigns.
Key Features and Functionality:
- Luna Co-Processor Family: Offers a range of AI solutions, including:
- Mentium LUNA-1: Delivers 2.4 TOPS performance at 70mW typical power, suitable for IoT applications.
- Mentium Luna-R1: Provides 2.4 TOPS performance at 75mW typical power, designed for space applications with radiation-hardened features and tolerance up to 200kRad.
- Mentium LUNA-2: Achieves 4.2 TOPS performance at 130mW typical power, ideal for IoT applications requiring higher performance.
- Mentium Luna-R2: Offers 4.2 TOPS performance at 135mW typical power, tailored for space applications with radiation-hardened features and tolerance up to 150kRad.
- Mentium SDK: A comprehensive toolchain supporting PyTorch, TensorFlow, and ONNX, facilitating the compilation, quantization, profiling, and deployment of deep-learning models onto Mentium's co-processors. It includes a high-fidelity emulator for performance validation prior to hardware deployment.
Primary Value and User Solutions:
Mentium Technologies addresses the critical need for efficient AI processing in environments where size, weight, and power constraints are paramount. By providing co-processors that deliver high-performance AI inference with minimal power consumption, Mentium enables real-time data processing and decision-making in applications ranging from space exploration to advanced security systems. This capability allows organizations to enhance operational efficiency, reduce energy consumption, and deploy sophisticated AI solutions in challenging and remote environments without extensive hardware modifications.
