ETRI releases open-source no-code tool for industrial AI development

Democratizing AI for the workplace

The Electronics and Telecommunications Research Institute (ETRI) recently announced the release of TANGO 2 as an open-source tool. This platform is designed to help people who lack specialized knowledge in artificial intelligence build their own AI models and services. The goal is to make it easier for workers on factory floors and other job sites to create practical tools for their specific environments.

Before this tool existed, creating an AI service was a multi-step process that required different types of experts. On-site workers had to prepare the data, AI specialists were needed to train the model, and software developers were required to integrate that model into actual systems. TANGO 2 changes this by consolidating all of these steps into a single, unified development process.

By automating the workflow, the institute aims to reduce the time required to launch an AI project. It also lessens the burden on companies to secure highly specialized personnel for every small task. Users can now select and run the specific functions they need directly through a web browser without having to write complex code.

From vision to generative AI

ETRI has been working on the TANGO project since 2021. The first version of the tool, TANGO 1, focused primarily on computer vision tasks like image recognition. The new iteration, TANGO 2, significantly expands these capabilities to include generative AI and large language models (LLMs).

TANGO 2 is specifically built to help industries manage and train LLMs using their own unique data. This allows businesses to create models that understand specific professional contexts rather than relying on general-purpose AI. The tool also includes a core technology called LLMOps, which is a solution for the development and operations of generative AI.

The institute has released the LLMOps technology on GitHub, which is the world's largest platform for open-source development. This move ensures that the underlying technology is accessible to a wider range of developers. By making these components open, ETRI is encouraging a broader ecosystem of innovation around the TANGO framework.

Deployment across diverse environments

One of the key features of TANGO 2 is its flexibility regarding where the finished AI model lives. Once a user builds and optimizes a model, they can deploy it across a variety of different environments. These options include large-scale cloud systems, private in-house corporate servers, and even individual local devices.

This versatility is important for industrial applications where data privacy or hardware limitations might dictate where a model can run. For example, a factory might prefer to keep its data on a local server rather than the cloud. TANGO 2 allows these users to maintain that control while still benefiting from modern AI capabilities.

The institute is currently working with various companies and research institutions to expand these applications. They are testing the technology in several sectors to see how it performs in real-world scenarios. These partnerships help refine the tool for practical use in diverse industries like manufacturing and medicine.

Real-world industrial applications

Several industries are already beginning to test TANGO technology for commercial use. In the shipbuilding and marine sector, a company called Avnautics is using it to create an autonomous navigation solution. This system analyzes data from various sensors to help ensure vessels operate safely in the water.

The agricultural sector is also seeing progress through Sure Data Lab. They are building a generative AI service that uses the knowledge of farming experts to create a tool for use on actual farms. Meanwhile, Seoul National University Hospital is using similar technology to automatically generate radiology reports from chest CT images.

In the medical field, the hospital plans to demonstrate a service that can predict cardiopulmonary and vascular diseases. These examples show how the tool can be adapted to highly specialized fields. Each application takes the general power of TANGO and applies it to a specific, high-stakes professional problem.

Commercialization and standards

ETRI is actively working to move TANGO technology from the lab into the commercial market. For instance, the company Persona AI received a transfer of TANGO technology to integrate into its AI agent platform, GEN STATION. This company is looking to combine AI agents with physical robots to create 'physical AI' solutions.

The institute is also coordinating with the Ministry of Science and ICT to facilitate these technology transfers. They are aiming to help companies innovate quickly by providing them with a foundation they can build upon. This strategy helps ensure that the research leads to actual products that people can buy and use.

Looking ahead, ETRI is pursuing international standardization for no-code LLMOps technology. They are working with the International Telecommunication Union's Telecommunication Standardization Sector (ITU-T) to reach final approval by May 2027. This effort aims to establish common standards for how these types of tools should function globally.