ETRI Unveils 'Tango 2' to Help Bring AI to the Mainstream Without Coding
IT DAILY ·
✦ AI Summary
ETRI unveiled the next-generation version of its open-source no-code framework, 'TANGO (Tango),' called 'Tango 2.'
Tango 2 expands the existing vision-centric MLOps framework to support no-code training and deployment pipelines for large language models (LLMs).
ETRI held the '5th 2026 Tango Community Conference' on the 11th at the Korea Science and Technology Hall in Gangnam-gu, Seoul, and shared the current status of related technology development.
BizCrushWant to follow Korean talks and live streams in real time?Try CapCrush on the web
The Electronics and Telecommunications Research Institute (ETRI) unveiled the next-generation version of TANGO, an open-source no-code framework that lets domain experts in industrial settings build and operate AI without coding, rather than professional developers. Through this, ETRI aims to lower the barriers to AI adoption across industries.
'Tango 2' expands the existing vision-centric MLOps scope. The expansion adds no-code support for large language model (LLM) training and deployment pipelines.
ETRI held the '5th 2026 Tango Community Conference' on the 11th at the Korea Science and Technology Hall in Gangnam-gu, Seoul. At the event, it shared the current status of related technology development.
At the event, Chang-sik Cho, ETRI's principal researcher and the Tango community leader, delivered the keynote under the theme, 'Introduction to the Tango LLMOps/MLOps Community.' In his presentation, Cho explained the framework architecture and demonstration results. The photo was credited to reporter Kwon Young-seok.
Tango is short for 'Target Adaptive No-code neural network Generation and Operation framework.' It was designed to support neural network generation, training, and operation for experts in traditional manufacturing and industrial fields. The intended users are workers in sectors that lack professional development staff, and they can use it through mouse clicks and guided pipelines without complex coding.
Tango began as a government project in 2021, and its full code has been public on GitHub since 2022. The researcher said the key to building AI models lies in data labeling and interpretation. He added that the right people for data labeling and interpretation are domain experts who know the field best. He went on to say that Tango is a tool that helps frontline practitioners directly control AI pipelines without developers.
Tango's core function is 'Target Aware,' which optimizes models to suit the target hardware. Target Aware reflects the constraints of the deployment environment and is applied to low-spec on-device deployment environments such as the cloud, Nvidia Jetson, and Raspberry Pi. This function automatically searches for the optimal neural network architecture (NAS).
ETRI is focusing on developing 'Tango 2' as the next step after 'Tango 1,' which concentrated on automatic vision model search. ETRI began expanding into the LLM area in July 2024. The goal of 'Tango 2' is to simplify complex LLM training pipelines and provide a runtime that covers on-device and physical AI environments.
At the same time, three demonstration cases in shipbuilding, agriculture, and healthcare were introduced. In the shipbuilding and maritime sector, the technology has shifted from simple vessel detection to LLM-based autonomous navigation collision-avoidance support. The shipbuilding and maritime technology learns navigation norms such as the International Regulations for Preventing Collisions at Sea. In smart farming, the LLM learned the tacit knowledge of a master tomato grower on irregular temperature and humidity control and implemented an automatic environmental control system.
In healthcare, ETRI is working with the Department of Radiology at Seoul National University Hospital. In this area, it is learning chest CT image data beyond X-ray lesion identification. It is currently demonstrating a medical support LLM model that helps draft radiology reports.
The released licensing and technology transfer policy was designed to avoid these constraints in a situation where specific foreign algorithms such as YOLO are widely used in the vision field and where there are commercialization license dependency issues with those algorithms. Accordingly, it basically maintains a Copyleft stance, while applying a dual-license policy in the form of technology transfer agreements for companies that want to commercialize the product.
The researcher said that in the case of Tango 1, it was in the position of catching up with existing global MLOps tools. He added that LLMOps still lacks an established framework standard, and said ETRI will pursue standardization through Tango 2 from an equal starting line with the global market.
Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=242848
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.
Keep reading
BizCrush