"We are using EXAONE to identify candidate substances that help prevent hair loss in a single day and are preparing to commercialize them. We also aim to cut the time from cancer diagnosis to personalized treatment design from more than four weeks to a single day to protect patients' 'golden time.'"
Im U-hyeong, co-head of LG AI Research, said this in a keynote speech at the "LG AI Talk Concert 2026" held on the 14th at the Convergence Hall in LG Science Park in Magok, Gangseo-gu, Seoul, unveiling examples of "expert AI" applied to science. On the day, LG AI Research disclosed expert AI cases in manufacturing, finance and science that it has built while solving a range of on-site industrial problems over the past six years.
First, in science, it achieved results by leveraging AI experts to quickly discover new candidate substances and reduce trial and error in the research process. LG AI Research, in collaboration with LG H&H, reviewed 420,000 candidate substances and identified the hair loss efficacy substance "Rhamsidil (Rhamsidil)" in one day. With D&D Pharmatech, it is developing a next-generation oral peptide drug.
Next, EXAONE 4.5 has been embedded in and is being used by the Ministery of Food and Drug Safety's new drug review AI. It is expected to play an important role in shortening future new drug approval review timelines and assisting reviewers' work.
An autonomous laboratory using AI is also under development. In this setup, AI reviews test results and independently designs and conducts the next experiment. Han Se-hee of LG AI Research said, "With a minimum number of experiments, it has become possible to explore more broadly and deeply," adding, "We expect to develop advanced materials needed in real industries more quickly using AI."
LG AI Research also introduced EXAONE Tabular, an AI expert foundation model for industrial sites. EXAONE Tabular understands and predicts process conditions and quality information. Even if process or product appearance images change, it can operate without retraining. With customized additional training, it also proved world-class capabilities.
After first being used at an LG Innotek site, EXAONE Tabular is reportedly drawing interest from global pharmaceutical companies and hospitals. LG projected that it could expand into fields for efficient operation of energy such as nuclear and hydro.
Lee Hwa-young, head of AI business development at LG AI Research, said, "Manufacturing process data coming from actual sites is vast and messy, with much to organize. We also have to look at the expense of adopting AI and what value AI creates when applied. We reduce uncertainty with small amounts of initial data and data acquired on site. We will build the most optimized AI to solve complex problems in industrial settings."
LG AI Research is also developing a vision inspection agent that autonomously performs data sampling, labeling and model training. Until now, whenever a new product was added to a process, it took months just to newly generate the data needed for vision inspection, but this has shortened that process. The vision inspection agent is slated to be deployed to inspection processes within the year.
Lee Sun-yeong, head of the Data Intelligence Lab at LG AI Research, said, "With industry-specialized AI foundation models, we can now respond flexibly to changing process conditions in industrial sites with frequent changes."
Examples of EXAONE BI (Business Intelligence) applied in finance were also mentioned. LG AI Research officially launched EXAONE BI, a finance-specialized AI solution that generates everything from data analysis to reasoning, forecasting and explanation, early this year. With London Stock Exchange Group (LSEG), it is expanding business in global markets, and with Koscom, it is targeting the domestic market.
LG AI Research said it will prove results by targeting broader markets beyond Korea, North America, Europe and the Middle East. It projected that the use of explainable and reliable predictive AI in financial markets will expand to a range of asset classes, including bonds and commodities.
Although the era of artificial general intelligence (AGI) has drawn near, LG has decided to focus on foundation models. It views AGI as still lacking competitiveness in terms of expense and value for solving real industrial problems. LG AI Research said that since its establishment in Dec. 2020, it has solved more than 100 accumulated intractable industrial challenges.
Co-head Im U-hyeong said, "Over the past six years, LG has enabled AI to understand industrial problems, and now we are preparing for a stage where AI makes its own judgments and takes action," adding, "Based on a robotics foundation model, we are moving toward a future of autonomous factories where the entire plant operates as a single intelligence, beyond automating individual robots."
The low AI adoption rate among domestic corporations was cited as a concern for business expansion. Head Lee Hwa-young said, "Domestic companies' AI adoption rate is low. When building EXAONE Tabular, we worked to make it lightweight so that it would deliver better-than-expected results while reducing GPU usage. Many overseas companies want to use it via cloud-based APIs, so we are considering various options."