Artificial Intelligence

Meltdown, Spectre, and other backdoor attacks exploit critical vulnerabilities and compromise data security. Malware threats each has its unique methods, typically stealthy and subtle to take over storage devices. In many cases, malware bypass prevention and execute encryption attacks followed by demands for payment from infected victims. Worst yet, even after the payment is made, data remains locked.

Computational storage offers real-time security scan combining storage and AI. We provide in-line AI neural network to 1) queries over petabytes of data 2) scalable real-time self-searching that offloads CPU 3) search engine optimized for intrusion detection 4) reduce latency such that petabytes of data can be aggregated with no increase in search time 5) a second layer of protection in case the primary layer in server CPU is compromised.   

We developed a scalable neural network in FPGA capable of perform training/classification with microseconds search time by applying massive parallelism . AI search engine inside storage such that petabytes of data can be aggregated resulting search time to reduce from days or hours to minutes.  

• FPGA Silicon-proven, high-performance AI/ML targeted for embedded low latency SoCs, ASICs and FPGAs.
• Solutions for facial recognition, automotive, security applications.
• AI/ML Inference for low latency eye tracking, automotive applications
• Unstructured dataset search / Cybersecurity applications
• Scalable neural processor with the RTL configurable parameters
• Optimized for training and inference processing, concurrent training and object recognition
• Support the AXI interfaces
• Ultra-low compact neural engine using embedded SRAM (no external DDR)