California-based Kinara is working with South Korean AI firm Enerzai to increase performance and reduce memory requirements for smart retail, smart city and industrial applications.

Kinara develops AI processors and a software development suite for edge computing applications. Enerzai compresses and optimises AI models for cost-effective and power-efficient edge AI applications.

While the Kinara Ara-1 Edge AI processor already delivers performance using AI models, the partnership with Enerzai helps joint users achieve a performance increase and reduction in memory requirements.

The Ara-1 Edge AI proprietary AI processor delivers a trifecta of performance, power efficiency and affordability, making it suitable for camera-based applications, spanning from smart retail to driver-monitoring systems, as well as smart city, drone and factory automation applications.

Ara-1 also features natural language processing capabilities, catering to the growing market for voice-controlled applications. Its development tools offer resources to convert neural network models into optimised computation graphs, all while streamlining deployment onto the Ara-1 chip.

“While our Enerzai model compression technology is architecture agnostic, Kinara’s AI processor provides a compelling advantage for customers that require an optimised combination of performance, energy and cost,” said Daniel Chang, Enerzai CEO. “To showcase this significant capability, we created a low-light enhancement demonstration that relates to fast camera, low-light image enhancement – a function that typically requires a very large and computationally intensive model. Running this demonstration on the Kinara processor, along with our model compression technology, created great synergy to significantly improve frame rates, which will satisfy our customers’ challenging edge AI requirements.”

Ravi Annavajjhala, CEO of Kinara, added: “The model optimisation capability of Enerzai technology provides a great complement to the Kinara Ara-1 AI processor by taking AI models and reducing the number of parameters and the computational loading. This combination provides our customers with the best of both worlds – a high performance AI processor and optimised AI models better targeted for resource-constrained edge AI applications.”

Kinara works with the TSMC foundry, and the company showed the results of its partnership with Enerzai at this week’s TSMC 2023 North America Technology Symposium in California.