dorsaVi Demonstrates Reflex AI with ≤0.8 Microseconds Latency and Ultra-Low Energy Use

Engineering Breakthrough in Reflex AI

dorsaVi Limited (ASX, DVL) has announced a significant technical validation of its RRAM-powered Reflex Platform, confirming its ability to perform neural reflex decisions at speeds and energy efficiencies comparable to biological reflexes. The company’s engineering analysis demonstrated that the platform can execute reflex decisions in less than 0.8 microseconds, consuming as little as 13 picojoules per operation. This represents a leap forward in enabling real-time, energy-efficient reflex control for robotics and wearable devices.

The breakthrough effectively removes the memory element as a bottleneck in reflex AI systems, shifting the challenge to optimizing peripheral components such as sensors, analog-to-digital converters, and actuators. This milestone is a critical step toward deploying reflex-grade artificial intelligence in practical applications where speed and power efficiency are paramount.

Neuromorphic Computing Meets Real-World Robotics

dorsaVi’s Reflex Platform leverages resistive random-access memory (RRAM) technology to perform neural network computations directly within the memory array, a method known as in-memory computing. The company implemented a compact multi-layer perceptron neural network designed for safety-critical tasks like distinguishing between “hold” and “tighten” commands in robotic grippers or prosthetic limbs.

This architecture enables reflex decisions to be made with sub-microsecond latency and ultra-low energy consumption, supporting continuous adaptation and recalibration without significant power drain. The platform’s energy efficiency and speed make it uniquely suited for battery-powered, always-on devices such as exosuits and human-machine interfaces (HMIs).

Overcoming Legacy Flash Memory Limitations

The announcement highlights the inadequacy of traditional flash memory technologies like NAND and NOR for reflex-class AI applications. NAND flash suffers from high latency and limited endurance for frequent writes, while NOR flash, despite faster reads, consumes too much energy and has limited write cycles. dorsaVi’s RRAM-based approach circumvents these issues by enabling fast, low-energy, and frequent neural updates directly within memory.

This positions the Reflex Platform as a pioneering solution for next-generation robotics and biomedical devices, where rapid reflexes and energy efficiency are critical for performance and safety.

Strategic Implications and Market Potential

Chairman Gernot Abl emphasized the significance of this validation, noting that it marks a pivotal milestone in dorsaVi’s transition from traditional sensor technologies to neuromorphic systems capable of real-time thinking and adaptation. The company is focusing on high-impact markets such as robotic grippers, prosthetics, exosuits, and HMIs, where the Reflex Platform’s capabilities can deliver tangible benefits.

With this engineering validation, dorsaVi is well-positioned to advance its Artemis Labs innovation program and move toward commercial deployment. The company also indicated that results from its ongoing robotic evaluation program will be reported imminently, which could provide further insights into practical applications and performance.