Theta Labs, the leading decentralized cloud for AI, media and entertainment, is pleased to announce that Data & Language Intelligence Lab at Yonsei University has signed a multi-year agreement to utilize Theta EdgeCloud across various AI research groups for AI model training and inference.
Theta EdgeCloud is the first hybrid cloud-edge computing AI platform, powered by the Theta Edge Network with over 30,000 distributed edge nodes, and cloud partners including Google Cloud and Amazon Web Services offering over 80 PetaFLOPS of always available distributed GPU compute power. This partnership opens up a significant market opportunity for Theta EdgeCloud in the academic research sector.
Yonsei University, ranked 1st in South Korea and 8th in Asia in the 2024 QS World Universities by prominent global higher education ranking institution, QS (Quacquarelli Symonds). Yonsei University Department of AI’s research areas include AI, machine learning, data mining, computer vision, natural language processing, robotics, and applied AI.
This partnership opens access to AI research by providing affordable, high-performance computing to academic institutions. Theta EdgeCloud enables university labs to compete with well-funded commercial labs while fostering greater transparency in AI development.
The collaboration between Yonsei University’s Department of AI and Theta will be led by Professor Dongha Lee, assistant professor at Yonsei University’s Department of AI and an adjunct professor at the graduate school of AI at POSTECH, a top South Korean research university.
Professor Dongha Lee's research focuses on knowledge enhancement for large language models, information retrieval, recommender systems, and data intelligence for real-world applications. He has made significant contributions to the field, including work on scalable machine learning systems and robust cross-lingual entity alignment. His innovative research has been widely published in prestigious conferences and journals.
This partnership follows on from Theta's recent multi-year agreement with KAIST AI, one of Korea's leading AI research labs. The collaboration, led by Professor Jaegul Choo at KAIST AI, focuses on leveraging Theta EdgeCloud for advanced research in computer vision and natural language processing. KAIST AI, often referred to as the "MIT of Korea," will utilize Theta's hybrid cloud-edge computing platform to accelerate their work on projects like StableVITON, a generative AI model for virtual clothing try-ons.
Yonsei University will utilize Theta EdgeCloud's GPU platform to research complex machine reasoning and data mining, leveraging the power of LLMs. This collaboration will enable PhD students to effortlessly train, fine-tune, and test AI models. Unlike commonly seen chatbot models that focus on understanding data in the form of natural language, this project will target extracting implicit knowledge and insights from tabular data by building a plug-and-play table reasoner and summarizer.
The intelligent table reasoner can self-question and answer it by faithfully pinpointing evidence on the table, providing explainable guidance for the summarizer. To train a reliable table reasoner, the team will collect table knowledge by guiding a teacher LLM to follow coarse-to-fine reasoning paths, refining it through multiple quality enhancement strategies to selectively distill high-quality knowledge to the reasoner. By utilizing the EdgeCloud platform instead of building in-lab GPU clusters, students can concentrate more on their research while significantly reducing costs for the university, thanks to Theta's competitive on-demand pricing compared to other major cloud GPU providers.
“The Data & Language Intelligence Lab at Yonsei University and I are thrilled to partner with Theta to utilize EdgeCloud for our AI model training and inference needs,” said Professor Dongha Lee. “Our current in-house GPU infrastructure has become costly to grow and scale, and requires too much time from our staff to manage and support. We’re finding Theta EdgeCloud’s hybrid GPU computing platform to be both scalable and flexible to meet our growth requirements while delivering the same service levels and performance compared to other cloud providers, at less than 50% of the cost. We’re looking forward to a successful long-term multi year partnership with Theta.”
“We’d like to welcome Prof Dongha Lee and his research lab at Yonsei as our second academia customer for Theta EdgeCloud,” said Mitch Liu, co-founder and CEO of Theta Labs. “Based in South Korea, Yonsei is one of the top AI research universities in the world and it’s a pleasure to work closely with Prof. Lee and his team of over 30 PhD/Masters students and faculty. We look forward to seeing their innovative AI initiatives and breakthrough advancements in data science, machine learning and other AI areas, all powered by the Theta EdgeCloud platform.”
