Matthew Wang is a co-founder of OpenGradient, a platform focused on decentralized infrastructure for hosting AI models and integrating machine learning with Web3 applications. He has a background in research engineering, quantitative modeling, and AI development for finance and blockchain technologies. ^[4]

Education

Wang graduated from Northwestern University with a degree in Electrical and Computer Engineering. ^[2]

Career

Wang began his career in 2018 as a software engineering intern at NASA, where he worked on preliminary hazard data analytics and modeling. In early 2019, he served as a software engineering intern at Meta, contributing to messaging heuristic infrastructure for Instagram and Messenger. Later in 2019, he joined Google as a machine learning engineering intern, where he worked on AI modeling infrastructure for the Google Ads traffic estimator. From 2020 to 2024, Wang worked as a research engineer at Two Sigma, where he conducted equity options market-making research. In 2024, he founded OpenGradient and served as its CEO, leading a research lab focused on research at the intersection of artificial intelligence and blockchain computing. ^{[1] [2]}

Interviews

NEAR Office Hours

In an interview with NEAR Protocol in December 2024, Wang discussed the development of OpenGradient and its aim to support decentralized infrastructure for hosting AI models, executing inferences, and deploying applications. He described the platform’s full-stack approach to integrating machine learning into Web3 applications, including a decentralized model hub that serves as an alternative to centralized AI model repositories. Wang outlined the underlying blockchain architecture, which relied on specialized nodes to handle transactions and AI inference efficiently, and explained how inference requests were verified using trusted execution environments and cryptographic proof mechanisms to ensure computational integrity. He also referenced use cases in areas such as decentralized finance and on-chain reputation systems. He described longer-term goals to expand AI adoption in Web3 through continued infrastructure development and research. ^[6]

Presentations

Adaptive AI Agents

In a presentation at ETHDenver in March 2025, Wang discussed the development of intelligent, adaptive AI agents. He outlined a progression from agents that synthesize information to those that execute tasks on behalf of users and ultimately to systems capable of managing complex operations autonomously. He identified current limitations in AI agents, particularly their difficulty in producing detailed analyses for complex financial and risk-related questions. He described key challenges in data access, computational workflows, and interoperability among agents. Wang explained how OpenGradient’s infrastructure was designed to address these constraints through a full-stack, on-chain approach that supported specialized nodes for inference and secure data pipelines. He also referenced early use cases in prediction markets and decentralized finance, demonstrated real-time analytical capabilities, and introduced a model hub intended to support the deployment of machine learning models for Web3 applications. ^[8]

Panels

AI Meets Web3

At Taipei Blockchain Week in January 2025, Wang participated in a panel alongside Ryuk of Iagent Protocol, Mark Rydon of Aethir, and Luki Song of Chainbase, discussing developments in AI and Web3. Wang described OpenGradient as a decentralized platform for hosting AI models, emphasizing secure integration for developers and verifiable workflows on blockchain, in contrast to traditional platforms like Hugging Face. The panel covered the roles of each organization: Iagent focused on visual learning infrastructure for gaming, Aethir provided decentralized GPU services, and OpenGradient supported secure AI/ML development. Discussions included target audiences, onboarding challenges for Web2 users entering Web3, approaches to data compliance and monetization in gaming, and the benefits of open-source AI models. The panel concluded with reflections on the broader potential of AI agents, including applications in gaming, portfolio management, and advanced visual learning. ^[5]

Future of Compute

At the Open AGI Summit in November 2024, Wang participated in a panel alongside Nick Emmons, Prashant Maurya, Jeremy Hazan, and Mikhail Avady, moderated by Brandon Potts, discussing centralized versus decentralized computing architectures. Panelists presented their respective projects, ranging from decentralized GPU networks to AI infrastructure designed to broaden access to computing resources. They emphasized that user needs, including performance, privacy, and cost, should guide the choice of architecture. They noted that centralized systems generally provide superior performance. At the same time, decentralized solutions offer cost efficiency and resistance to censorship, but acknowledged that reliability and usability challenges have led some users to revert to traditional cloud services. The discussion concluded on an optimistic note about emerging technologies, such as verification mechanisms and privacy-enhancing tools, as drivers of wider adoption of decentralized computing. ^[9]

New Era for AI

In November 2024, Wang participated in a panel at NEAR Protocol’s REDACTED conference alongside Mark Rydon of Aethir, Mathilda Sun of Gaib, and Cameron Dennis of NEAR, focusing on the future of decentralized computing. Wang discussed his transition from quantitative modeling to decentralized systems, highlighting OpenGradient’s efforts to provide out-of-the-box solutions for hosting and integrating compute into applications, including a staggered pricing model for GPUs, TEEs, and CPUs. The panel also covered the roles of other participants, including Gaib’s financial layer for GPU-backed assets and Aethir’s enterprise-grade decentralized GPU infrastructure for AI and gaming, as well as community concerns over latency and service reliability. Wang provided examples of practical applications, such as optimizing automated market-maker trading fees and lending-protocol risk models, and addressed regulatory considerations affecting decentralized computing. He concluded with insights into the potential of autonomous agents leveraging permissionless compute to perform complex machine-learning–driven tasks. ^[7]