GPU-based AI has become an indisputable trend in the technology industry. The market has been swamped by a stream of AI-based smart services such as smart transportation, smart medicine, and self-driving cars. Many research institutes are also encouraging enterprises to adopt AI technology in order to accelerate their internal digital transformation process. With GPU performance levels rising rapidly and open-source AI algorithms more accessible now than ever, enterprises can enjoy much lower entry thresholds into the field of AI than ever before. However, it is also true that many challenges still remain and that not all enterprises can transition smoothly into the age of GPU-based AI.
Statistics show that enterprises hoping to integrate AI technology will likely face the following obstacles: lack of human resources, limited availability of computation resources, and time constraints. Currently, companies (including major IT companies) must compete fiercely to hire software engineers skilled in AI talents. There is a shortage of engineers for AI teams. Typically, data scientists are too overwhelmed with other work to focus on AI application development. Secondly, computation resources are expensive, development environment configuration is complicated, and optimization can be a daunting task. Finally, training AI models is both costly and time-consuming. Such lengthy training times often cause projects to take exceedingly long periods of time. Given such circumstances, it typically takes at least one year for enterprises to complete AI-related projects, while cases that take over two years are also extremely common. The hAIsten AI service from Avalanche Computing is designed to tackle the aforementioned issues to help enterprises quickly access AI-driven results.
According to a study published by AI 100 in 2020, more than 70 percent of AI developers have to spend a lot of time building AI application environments and debugging development environments, said Jay Chen, Avalanche Computing CEO. To combat this problem, Avalanche Computing has introduced hAIsten AI. This software service allows AI developers to focus on the AI model design process rather than assembling hardware, setting up software environments, or optimizing execution environments.
Accelerating the commercialization of products
The team of Avalanche Computing is made up of engineers that formerly worked at NVIDIA's pursuit engineering solution architect (PESA) team. This has given the company an exceptional edge on deep learning technology and supermassive computation. The hAIsten AI service provided by the company uses a software-as-a-service (SaaS) model that acts as a low-code AI development platform that is both fast and easy to use. Domain experts and AI engineers in various fields from different enterprises can all access one intuitive AI computing environment directly without having to work with complicated code. They can simply use the optimized AI computation environment to quickly and efficiently finish the training and deployment of their AI models. By doing so, enterprises can skip the AI environment construction and system optimization processes. Furthermore, hAIsten AI can automatically access multiple GPUs and combine their computing resources for AI model training, significantly reducing training times. The one-click large scalable deployment function allows for the simultaneous deployment of various AI models to a large number of devices using a graphical user interface to facilitate not only AI development, but also application deployment.
In other words, hAIsten AI allows domain experts and AI developers to focus on their professional work, dramatically increasing their productivity, making their programming workflows easier. This innovative technology has also made Avalanche Computing one of the 10 finalists in the Qualcomm Innovate in Taiwan Challenge (QITC) 2021 and one of Hello Tomorrow's Deep Tech Pioneers. The company has also been selected to join world renowned Berkeley SkyDeck Global Innovation Partners (GIP) Program with the support of the TTA team in Silicon Valley.
Thanks to the prevalence of public cloud services, enterprise users can circumvent the heavy burden of building AI computation hardware by renting GPU resources. But even with rented GPU power, debugging and optimizing the AI computing environment, AI model design, AI model training, and deployment are time-consuming tasks that can delay projects. Jay stated that the goal of hAIsten AI is to remove these troublesome tasks to let professionals focus on more meaningful work, so that computation resources can be effectively utilized to turn innovation into business models. Avalanche Computing has already entered into partnerships with manufacturing and medical vendors, and will continue to expand the applications of the company's software service to sow more seeds of success.
Avalanche Computing, for instance, has worked with a medical startup that focuses on providing breathing monitoring systems for medical professionals. The startup's smart stethoscope combines professional recording equipment with AI to give healthcare professionals the most accurate real-time information. In order to achieve accurate predictions, the team's data scientists had to repeatedly try different AI algorithms and adjust AI model parameters, which inevitably lengthened the product development cycle. Luckily, after introducing Avalanche Computing's hAIsten AI software, they were able to overcome many of these problems. The startup also decided to move the development of their AI model to a cloud platform with hAIsten AI. hAIsten AI was then able to utilize multiple GPUs for the training process to validate the performance of each algorithm in much lesser time. In the end, data showed that the startup was able to accelerate their AI model development speed by more than three times, helping them to promptly seize market opportunities.
As AI becomes more and more prevalent, Avalanche Computing plans to keep working with its existing Taiwanese customer base while striving for more exposure and fundraising opportunities to enter the American, Japanese, and European markets, among others, to create better living for people around the globe through AI.
hAIsten AI
Photo: Avalanche Computing
(Editor's note: The original article was published in TTA Magazine Issue 9. Read more startup stories in TTA Magazines.)
