Making Deep Learning Accessible: Dreaming a New Standard for Deep Learning Software

[Hongsuk Lee, CEO of Neurocle]

Dreaming of AI deep learning accessible for those without AI expertise

The emergence of AlphaGo raised awareness around the potential of AI, and in recent years AI has become a current issue rather than a distant future. Neurocle focused on the demand for artificial intelligence and started with the idea of developing deep learning software for domain professionals to use directly in any industry without deep learning experts.

Hongsuk Lee built a team of people with industry experience from large companies and made product launch his top priority. In less than a year, they launched Neuro-T version 1.0 and succeeded in seeing twice the growth every year.

Deep Learning SW for AI Non-Professionals, Neuro-T & Neuro-R

Neuro-T and Neuro-R are GUI-based deep learning vision learning software that allows anyone to use deep learning vision technology.

As a general-purpose software that enables non-AI experts to create and utilize models for image analysis based on deep learning vision technology, Neuro-T and Neuro-R provides the Auto Deep Learning algorithm and a systematic learning data management system through which deep learning models are automatically generated with just a few clicks.

The T in Neuro-T stands for training, meaning a GUI-based software for training and creating AI deep learning models, while the R in Neuro-R stands for real-time, meaning an inference optimization API, which can be integrated on various systems and can be run in real-time onsite.

Universal Artificial Intelligence SW for Any Industry

Most companies developing AI target specific industries, whereas Neurocle initially aimed to develop a solution that could be applied to all sectors.

Although Neuro-T and Neuro-R are still in their early stages, they are being used in a variety of industries such as manufacturing, healthcare, logistics, and security. Through Neuro-T and Neuro-R, defect detection of manufacturing processes, analysis of medical image data, logistics automation, and analysis of security footage is incorporating artificial intelligence technology.

This advantage of being able to easily incorporate AI into any field has led to unexpected but favorable outcomes in education. As AI technology becomes commonplace, Neurocle software is being used as educational tools for learning deep learning in high schools and universities. The fact that students without any previous background in deep learning can create deep learning models through Neuro-T is an example of a step closer to the vision of "Making Deep Learning Vision Technology More Accessible."

The Key is Data-centric Algorithm Optimization

With Auto Deep Learning technology, deep learning is applicable to any field even if the user is not an AI expert. Depending on the nature of the input data, results can be produced at a level similar to that of direct optimization performed by an expert. Neurocle spent a good amount of the development phase to carefully craft the algorithm optimization, and test results show similar or better results when compared to domain-specific algorithms. Based on this technology, Neurocle is a “First-Mover” in the field of Automated Deep Learning.

Technology and Sales Strategies Work Together to Break-even in 3 years

Many startups struggle to cross the break-even point and commercialize their products. As a result, most startups raise money from external investment to recruit employees and run their business. However, Neurocle didn't follow this typical route. By launching the product in its first year, sales have increased by more than twice annually for three years. This year, three years since its founding, Neurocle surpassed the break-even point on its own. In just three years, this is quite an achievement achieved without national R&D support or investment from private investment companies.

Along with excellent technical skills, the decision to launch the product within a year and the individualized sales methods for industries made a difference. Rather than persuading buyers solely on the technical superiority of the developed products, Neurocle put forward different sales strategies according to the value chain of each industry and the characteristics of the players. For the sophisticated value chain of the manufacturing sector, Neurocle targeted SI companies, while targeting end users, university hospitals and research institutes for the medical/bio sector. Internationally, Neurocle sells its products through local agencies.

Named 2021 Gartner Cool Vendor for Driving Innovation to Advance Deep Learning

In 2021, Neurocle won the Korea Association for ICT Promotion (KAIT) President's Award at the 2021 Korea ImpaCT-ech Awards. In the same year, Neurocle received the Vision System Design Innovators Awards and the Korea Digital Education Frontiers Association (KEFA) President's Award at the 6th E-Learning·Edutech Business Model Competition. At each event, Neurocle was accredited for the technology that allows non-experts to create and utilize high-performance deep learning vision models.

In particular, the selection as 2021 Cool Vendors in AI for Computer Vision by Gartner, the most influential global market research firm, is a significant achievement for Neurocle as it prepares for full scale global market entry. Neurocle has been accelerating its expansion abroad, collaborating with various institutions including Denso and Panasonic, as well as participating in the German exhibition last October. CEO Hongsuk Lee disclosed, "As we expect global companies to refer to Gartner's report, we take this as an opportunity to actively target global markets."

Gartner, a global IT research firm, analyzes creative, innovative vendors (suppliers) leading the future and selects comprehensively excellent companies to introduce them on their 'Cool Vendor' report. Gartner's report is cited in media around the world, and many companies pay attention to their report to predict technology trends and use the evaluations of companies and products for business.

High-performance Computing Fund Increases Productivity By 5%

As much as deep learning optimization is important, computing resources are crucial for the testing and validation phase of Neurocle's R&D. This year, the resources supported by the Artificial Intelligence High Performance Computing Support Fund from the National IT Industry Promotion Agency were used to experiment and validate our products. "In AI research and development, resource performance affects model performance. We appreciated the national support, which reduced development time and reduced costs, resulting in productivity gains of about 5 percent," said CEO Hongsuk Lee.

Becoming the Adobe of Deep Learning

CEO Hongsuk Lee, who had set specific goals from the beginning of the company, also gave a clear answer for his future plans. Ultimately, what Neurocle dreams of is to become a company like Adobe in the field of artificial intelligence. Adobe is a global company that offers a standard SW that is used by both professionals and amateurs alike in the field of computer graphics. Neurocle aims to be a company that provides standard software in the field of artificial intelligence. In the future, the adoption of artificial intelligence will be accelerated regardless of industry or company size. In this process, Neurocle aspires to provide services that make it easier for anyone to take advantage of high-quality deep learning technologies. Neuro-T and Neuro-R, which were first released in 2019, were released as version 2.0 this summer. Through continuous development, optimization, and subscription-based updates, we want to make high-quality deep learning available to everyone.

Neurocle's goal is not to be biased towards external growth, but to become a company that makes substantial profits along with revenue growth. Lee stated that Neurocle has grown on its own without external investment in order to give employees more stock options. A company where employees grow together is a what Neurocle dreams to become.