What is OpenGradient, the open-source AI platform incubated by a16z's startup accelerator and favored by Balaji?

Compiled by: Scof, ChainCatcher

Edited by: Nianqing, ChainCatcher

Today, the centralized open-source AI platform OpenGradient announced the completion of a $8.5 million seed round financing. Investors include a16z Crypto Startup Accelerator, Foresight Ventures, SV Angel, Coinbase Ventures, SALT Fund, and Symbolic Capital. Well-known industry figures such as angel investor Balaji Srinivasan, NEAR founder Illia Polosukhin, and Polygon founder Sandeep Nailwal also participated as angel investors.

The funds will be used to continue building decentralized infrastructure, deploying solutions and tools for AI and Web3 developers, and advancing applied ML research to develop an open-source AI ecosystem on the blockchain. OpenGradient's testnet is expected to be open to developers in Q4 2024.

In early September, OpenGradient was also selected for the a16z Crypto Fall Crypto Startup Accelerator (CSX) program. a16z CSX invests at least $500,000 in each selected company, and the program will provide 8 weeks of training in New York City, offering firsthand experience from top crypto founders, guidance and support from the a16z crypto investment and operations team, and connections with other companies and the broader a16z network.

According to the official website, OpenGradient is the first decentralized platform for hosting open-source AI models, secure execution, agent inference, and application deployment. OpenGradient's novel Heterogeneous AI Computing Architecture (HACA) enables permissionless AI inference, statistical analysis, data processing, and agent execution to run securely and at scale end-to-end via smart contracts, while building a web platform designed to accelerate AI research. OpenGradient allows developers to seamlessly leverage open-source models and utilize computing for risk management, ML optimization, AI agents, anti-sybil, DeFi mechanism design, and more.

The following is compiled from the OpenGradient official blog, with some content edited and supplemented:

The Fusion of AI and Blockchain

Artificial intelligence has become an indispensable part of our daily lives: from powering smartphone assistants and recommendation algorithms to providing advice through conversational robo-advisors for stock recommendations, AI impacts nearly every aspect of modern society. However, the current state of AI is not only centralized, controlled by a few tech giants that can arbitrarily restrict access to AI models, but its computational processes also lack sufficient guarantees and security, making it unsuitable for applications that require high verifiability, such as blockchain applications. Therefore, while AI is ubiquitous in enhancing traditional applications, Web3 protocols lag behind Web2 in complexity, optimization, and user experience.

Major obstacles to integrating AI into Web3:

• Blockchain is limited by virtual machine constraints and computational capacity
• Off-chain inference is trusted but vulnerable to attacks
• The hardware requirements and computational costs for hosting and servicing AI models are high
• Custom infrastructure needs to be built for trust-minimized inference (via zkML, trusted execution environments, etc.), which is complex
• Lack of research exploring the design space for AI-driven Web3 applications

The Emergence of OpenGradient

OpenGradient is the solution that merges blockchain with artificial intelligence to address their respective challenges. OpenGradient is the first permissionless platform for open-source model hosting, secure inference, agent inference, and application deployment. By developing AI tools, feature-rich libraries, and an innovative decentralized platform based on blockchain, OpenGradient not only makes the integration of AI workflows into Web3 applications secure and seamless but also decentralizes and democratizes access to AI models.

OpenGradient is building an EVM-compatible blockchain network aimed at becoming a scalable and secure execution layer for AI. The network employs a revolutionary approach to reshape blockchain architecture, allowing developers to infer AI models directly from their Solidity smart contracts through simple function calls.

Since OpenGradient Network provides direct access to inference AI models through precompiles in smart contracts, it can maximize the composability of smart contracts.

Core products and plans include:

• Web3 AI Research - OpenGradient's research team drives innovation in open-source AI in the crypto space by building models for DeFi risk analysis, AMM dynamic fee optimization, DePIN reputation, and more.
• Model Hub - OpenGradient's model hub is the Web3 version of HuggingFace: an easy-to-use web portal where any developer can upload, download, or access any AI model in our decentralized and censorship-resistant file storage.
• OG SDK - The OG Python SDK provides seamless access to OpenGradient's functionalities, including model access, secure inference, proof/certification retrieval, and more.
• NeuroML - NeuroML allows developers to leverage models for ML optimization, systemic risk management, AI agents, witch defense, DeFi mechanism design, and more, seamlessly building within our AI application ecosystem.
• HACA Infrastructure - OpenGradient's proprietary Hybrid AI Computing Architecture (HACA) expands and protects on-chain AI workflows through node specialization, providing inference, agent inference, statistical analysis, and more.

Team Information

OpenGradient is headquartered in New York City, and the team consists of global talent with deep expertise in AI/ML, blockchain, cryptocurrency, and the Web3 ecosystem. Team members come from tech companies such as Google, Coinbase, Ripple, Intel, and Palantir.

Co-founder and CEO Matthew Wang holds a Bachelor's degree in Computer Engineering from Northwestern University and is a research engineer at Two Sigma, with experience in quantitative modeling and machine learning research.

Co-founder and CTO Adam Balogh was the head of the AI platform at Palantir Technologies and has 7 years of industry experience. He holds a Master's degree in Advanced Computing from Imperial College London.