# LG says Exaone now finds new materials, picks stocks for clients > Source: > Published: 2026-07-08 04:15:31+00:00 LG AI Research put its Exaone model's industrial results on display at the International Conference on Machine Learning in Seoul this week, presenting 14 papers alongside AI-discovered industrial materials, a stock analysis agent and a data generation platform already in commercial use, the company said Wednesday. It is the first time the conference has been held in Korea. ICML, one of the three most prominent global AI and machine learning conferences, runs Monday through Saturday at Coex in southern Seoul. LG's exhibit centered on new materials. The company displayed samples of Rhamsydil, a hair loss treatment ingredient, and an immersion cooling fluid for AI data centers, both discovered through its Exaone Discovery platform. Rhamsydil was found with LG Household & Health Care from more than 420,000 candidate compounds, screened by AI in a single day. It works without steroid-derived components, according to results LG presented at the World Congress for Hair Research, and is being prepared for commercialization. The cooling fluid was co-developed with GS Caltex, and the two companies plan to expand their joint materials work. Exaone Discovery reads molecular structures from scientific literature and designs candidate compounds to hit specified performance targets. LG registered a patent covering the process earlier this year and applies it to cosmetics, batteries, semiconductor materials and drug research. In finance, LG demonstrated Exaone BI, an AI agent that analyzes about 8,000 stocks listed in Korea and the US daily, generating prediction scores and analyst-level commentary. LG launched a US market prediction service with the London Stock Exchange Group earlier this year and signed a contract with Koscom this week to begin a Korean equivalent. A third platform, Exaone Data Foundry, generates training data and automatically builds domain-specific models. LG says it raises data productivity at least a thousandfold and improves quality by an average of more than 20 percent. In a pilot with the National Pension Service, the system built over 10,000 specialized data records a day. "Exaone is no longer an experiment; it's an expert AI that finds answers on site," said Lim Woo-hyung, co-CEO of LG AI Research. "We'll take the lead in global AI through a first mover strategy aimed at solving hard industrial problems." Among the 14 papers, LG highlighted its generative AI for new materials, which placed second worldwide on LeMat-GenBench, a benchmark that scores the stability, novelty and diversity of AI-generated crystal structures. Since launching in December 2020, LG AI Research has published 363 papers at top-tier AI conferences and filed 838 patent applications, including 371 in Korea, 243 overseas and 224 through the international PCT system. The company also used the conference to recruit, hosting about 60 master's and doctoral researchers attending ICML at an "LG AI Day" networking event Tuesday. "AI leadership comes from the people who build the best technology," said co-CEO Lee Hong-rak. "We'll build a system that gives researchers the best experimental environment, data and mentoring, and become a top global research hub." mjh@heraldcorp.com