dorsaVi Launches Modular Ultra-Edge Hardware Platform for Robotics and AI
dorsaVi has initiated its Ultra-Edge Modular Design program, transitioning from IP development to a manufacturable hardware platform targeting robotics, exoskeletons, and industrial AI with ultra-low power consumption.
- Modular hardware separates sensing, compute, memory layers
- Sub-1mW power architecture for coin-cell battery devices
- API layer enables integration with robotics and IoT partners
- Program guided by industry experts including ex-Omron VP
- Targets large markets like autonomous vehicles and industrial AI
Transition from Lab IP to Manufacturable Hardware
dorsaVi Limited (ASX:DVL) has taken a significant step beyond research with the launch of its Ultra-Edge Modular Design and Build program. This marks the company’s move from technology validation to creating a physical, manufacturable hardware platform designed for real-world deployment in robotics, exoskeletons, and industrial AI applications. The program is the third phase in dorsaVi’s five-stage Execution Plan and aims to transform its proprietary RRAM semiconductor and neuromorphic computing IP into partner-ready products.
This transition is pivotal as dorsaVi has spent recent months proving the synergy of its RRAM memory and neuromorphic processing technologies, which was highlighted in its recent RRAM-Neuromorphic Platform validation. The new program will now focus on turning these validated technologies into hardware that can be embedded into dorsaVi’s existing FDA-cleared sensor products and licensed to external partners.
Modular Architecture Enables Flexibility and Scalability
At the heart of the program is a three-layer modular hardware architecture that cleanly separates sensing, compute, and memory functions. This design allows dorsaVi to configure and upgrade each layer independently, providing the flexibility to tailor solutions for a range of applications without redesigning the entire system. The sensing layer leverages dorsaVi’s clinically proven wearable sensor technology, the compute layer utilizes a neuromorphic “reflex engine” for brain-inspired, ultra-low-power processing, and the memory layer incorporates RRAM technology developed in partnership with NTU Singapore and ITRI.
This modular approach not only addresses technical challenges but also opens diverse commercial pathways. dorsaVi can license or partner at the individual layer level, expanding its reach across different verticals such as robotics, autonomous systems, and clinical wearables. Such a strategy contrasts with traditional monolithic chip designs that limit adaptability and commercial models.
Power Architecture Targets Sub-1mW Operation
A standout feature of the program is its power architecture, engineered specifically for in-memory compute operation within a sub-1mW power budget. This ultra-low power consumption is essential for coin-cell battery-operated devices and always-on autonomous sensors; markets where conventional Von Neumann architectures fall short. By embedding AI inference directly within the RRAM memory array, dorsaVi aims to overcome the energy inefficiencies of data shuttling between memory and compute units.
The power management system includes voltage regulation, power gating for near-zero power states during inactivity, and compatibility with energy harvesting for remote or embedded applications. This technical innovation aligns with the growing demand for edge AI devices that operate reliably without cloud connectivity or frequent battery changes.
API Layer to Simplify Partner Integration
dorsaVi is concurrently developing an API layer to provide a clean, documented interface for embedding its ultra-edge intelligence stack into partner platforms. This integration interface is designed to abstract the complexity of RRAM physics and neuromorphic architecture, enabling robotics manufacturers, industrial automation integrators, and autonomous system developers to adopt dorsaVi’s technology without deep semiconductor expertise.
The API strategy is crucial for commercial scalability, allowing dorsaVi to participate in the robotics and industrial AI markets as a platform provider rather than building every end application itself. This approach reflects a broader trend in advanced hardware sectors, where proprietary technology layers are exposed to partner ecosystems to extend market reach.
Industry Input Shapes Design for Next-Generation Robotics
- Collaborative robots demanding sub-millisecond, cloud-independent safety decisions
- Autonomous drones operating on milliwatt power budgets without GPS or connectivity
- Industrial automation systems requiring real-time anomaly detection without cloud latency
- Exoskeletons and prosthetics needing instant response on wearable, battery-efficient devices
These needs map precisely to dorsaVi’s modular design, power efficiency, and neuromorphic compute capabilities, positioning the company to address large markets projected to reach hundreds of billions of dollars by 2030, including autonomous vehicles (USD 214B), robotics (USD 218B), and industrial IoT edge hardware (USD 68.7B).
Bridging Lab Success to Commercial Reality
dorsaVi acknowledges the common pitfall where promising chip technologies fail to become manufacturable products. Its modular strategy is a deliberate response, allowing independent testing and refinement of each hardware layer before full system integration. The company plans to validate the platform initially within its own FDA-cleared sensor hardware, ensuring regulatory compliance and real-world performance data before broader partner deployment.
This staged approach also strengthens dorsaVi’s IP position, as it controls each layer of the stack independently, enabling flexible licensing and partnerships. The company’s prior work on wafer-level RRAM testing and firmware upgrades adding on-sensor intelligence laid the groundwork for this hardware platform, as noted in its recent RRAM Testing for AI Memory and Firmware Upgrade for On-Sensor Intelligence.
CEO Mathew Regan emphasised that the modular design balances early development speed with long-term flexibility, allowing dorsaVi to adapt to diverse commercial pathways across robotics and autonomous systems. The integration interface is equally vital, facilitating sub-millisecond local inference on highly constrained power budgets without requiring partners to develop semiconductor expertise.
Bottom Line?
dorsaVi’s modular hardware platform could redefine ultra-low-power edge AI integration, but commercial adoption hinges on successful validation and partner uptake.
Questions in the middle?
- How quickly will dorsaVi validate and commercialise the full modular platform?
- Which partners or sectors will first adopt dorsaVi’s API-enabled ultra-edge modules?
- Can dorsaVi’s modular approach outpace competitors in the rapidly evolving edge AI hardware market?
