Saratoga, CA, December 18, 2020 – Aizip, Inc., a startup company focusing on artificial intelligence (AI) for applications in the Internet of Things (IoT), announced today the operation of its automated tiny machine learning (TinyML) design tools.
Traditional deep neural network (DNN) design involves the manual adjustment of many design factors. This process usually requires a team of engineers to expend a significant effort over an extended period of time. The results may not be fully optimized due to the limited AI talents available. For DNN in IoT applications – that is, TinyML – there are additional challenges in compilation and runtime due to the hardware variations and constraints. These problems make it extremely difficult to build a scalable business in the area of TinyML design.
A set of automated design tools is necessary to solve these problems. Aizip has successfully developed such tools for its design applications. These tools are highly automated for network architecture search (NAS), quantization-aware training (QAT), and transfer learning, which yield high-performance DNN within a short period of development time. Aizip’s NAS is implemented with quantization awareness, which makes it ideal for the TinyML market. Aizip’s transfer learning emphasizes the robustness of the model, which provides stable performance across different situations. Another feature is consistency under various design targets and hardware parameters. Fast turn-around time, which is made possible by the automated tools, is critical for customers in the dynamic IoT space. Multiple patents have been filed in relation to the significant innovations Aizip has made along the way.
“We’re very excited with the readiness of our automated design tools,” commented Yuan Lu, President and Co-Founder of Aizip. He further discussed the impact on Aizip design service: “Now we can claim a leadership position in the fast-growing AIoT industry. Our customers will get superior TinyML models quickly with great cost savings. Their return on investment (ROI) will be significantly improved.” Lu, who has extensive expertise in AI, IC, and design automation, received his Ph.D. from Carnegie Mellon University. The design tools have been successfully tested on Aizip projects with customers in the U.S. and overseas.
Lu attributes the rapid development of the automated design tools to Aizip’s outstanding technical team and smooth engineering processes. The engineering team covers both DNN design and implementation areas, with a deep understanding of the entire design process. Equally important is an efficient engineering operation process. This is made possible by a having a seasoned management team with many years of experience. This set of automated design tools will continue to be further optimized and expanded to its automated design platform.
The automated TinyML design tools will allow Aizip to serve many more customers with a scalable model. Aizip is currently accelerating its engagement with customers. For more information, please contact info@aizip.ai.
* * *
About Aizip, Inc.
Aizip, Inc., develops AI models for edge applications. Based in Silicon Valley, Aizip provides models with superior performance, quick turnaround time, and excellent ROI. These models can be used in a wide range of applications for an intelligent, automated, and connected world. For more information, visit www.aizip.ai.