One of our core missions at Theta Labs has been to democratize access to AI compute. Academic research labs are doing some of the most important & interesting work in AI today, yet they consistently face a structural disadvantage compared to commercial labs: access to affordable, high-performance GPU infrastructure (we’ve written about this problem at length here).
A PhD team at a top university simply cannot compete with the reserved compute clusters of a Google or Microsoft. The biggest thing we can offer them is a way to close that gap.
Over the past two years, Theta EdgeCloud has become the GPU platform of choice for a growing number of the world’s leading research institutions. What follows is a summary of the research currently taking place on our network, and the academics driving it.
KAIST, South Korea
KAIST ranks fifth in the world for AI research output at the top machine learning conferences, placing it behind only CMU, Stanford, UC Berkeley, and MIT. Professor Jaegul Choo’s Davian Lab was our first academic customer, joining in July 2024 to train next-generation computer vision models on EdgeCloud’s A100 and H100 clusters. Their work on StableVITON, a generative AI model for virtual try-on, was presented at IEEE CVPR 2024 and has direct commercial applications in e-commerce and fashion.
Yonsei University, South Korea
Ranked first in South Korea and eighth in Asia by QS, Yonsei University’s Data and Language Intelligence Lab joined in August 2024 under Professor Dongha Lee. The lab is using EdgeCloud to build a plug-and-play table reasoning system that extracts implicit knowledge from structured tabular data using large language models, with applications in finance, healthcare, and scientific research. Professor Lee noted that EdgeCloud delivers comparable performance to other cloud providers at less than half the cost.
Seoul National University, South Korea
South Korea’s top national university joined in November 2024, with the AIoT Lab led by Associate Professor Hyung-Sin Kim, a former postdoctoral researcher at UC Berkeley with over 70 publications and six U.S. patents. The lab is using EdgeCloud to advance healthcare AI including real-time on-device sleep monitoring, federated learning systems that preserve patient data privacy, and mixed reality applications. Their research has appeared at CVPR, ICCV, and ACM MobiSys.
University of Oregon, United States
Oregon’s Distopia Laboratory signed a multi-year agreement in October 2024, led by Assistant Professor Suyash Gupta, whose paper on Byzantine fault-tolerant consensus won the Best Paper Award at EuroSys 2023. The lab bridges distributed systems, blockchain, and federated learning, using EdgeCloud to run large-scale AI model training across parallelizable workloads that would be prohibitively slow or expensive on traditional cloud infrastructure.
Imperial College London, United Kingdom
Imperial became our first European academic partner in January 2026. The Security and Machine Learning Lab, led by Dr. Sergio Maffeis, conducts research at the intersection of AI and cybersecurity, including the CLOUSEAU framework, a multi-agent system that automatically investigates and explains cyberattacks, presented at ACSAC 2025. The lab uses EdgeCloud’s hybrid infrastructure to access community RTX nodes for prototyping and enterprise A100 and H200 clusters for large-scale training, without the overhead of managing separate contracts.
Syracuse University, United States
Assistant Professor Junzhe Zhang joined in July 2025, bringing research in Causal Reinforcement Learning, an emerging field that integrates causal inference theory with sequential decision-making. His lab conducts large-scale experiments in counterfactual reasoning, off-policy learning, and AI fairness on EdgeCloud’s hybrid infrastructure, which uniquely combines community-run GPUs with high-end A100s, H100s, and AWS Trainium and Inferentia instances. Professor Zhang described Theta EdgeCloud as the only platform offering that level of flexibility without compromise.
Soongsil University, South Korea
The most recent addition to our network, joining in June 2026, Soongsil’s HUMANE Lab is led by Professor Kunwoo Park and focuses on large language model research with a human-centered mission: tackling misinformation, bias, and fairness in AI systems. The lab won first place at the FEVER 2026 fact verification challenge at EACL, building a multimodal system that verifies image-text claims across four coordinated AI pipeline stages. Their work is grounded in open LLMs, and the philosophical alignment with Theta’s open, decentralized infrastructure made the collaboration a natural one.
Building a Global Research Network
These seven institutions represent a cross-section of a network that now spans 36 academic partners worldwide, including Stanford University, NTU Singapore, Brandeis University, and City St George’s University of London, among many others. They cover computer vision, speech recognition, healthcare AI, cybersecurity, causal inference, and responsible AI, all running on Theta EdgeCloud’s distributed GPU infrastructure.
If you lead an academic research lab and would like to discuss access to Theta EdgeCloud’s GPU infrastructure, reach out to us at [email protected].
