One of the biggest puzzles in the exciting world of blockchain is figuring out how to handle complex computations and massive amounts of data securely, all while keeping everything verifiable. Blockchains, while fantastic for secure and transparent transactions, weren’t originally built for heavy-duty parallel processing, especially the kind needed for cutting-edge AI. They excel at ensuring actions are predictable and agreed upon by everyone, but not necessarily speed and data crunching.

To tackle this challenge head-on, from March 7th to 9th, bright student developers and researchers from UC Berkeley, UWaterloo, and ETH Zurich teamed up for the Flare x Google Cloud hackathon. Organized in collaboration with Blockchain at Berkeley, the event was all about exploring how blockchains could power AI-driven computations without sacrificing that crucial element of verifiability.

Participants got hands-on with Google Cloud Confidential Space, a super-secure environment, like a digital vault, that proves its security through cryptographic attestations. It uses Trusted Execution Environments (TEEs) to ensure hardware-level confidentiality. The beauty is that the Flare blockchain can then check these attestations. This verification process guarantees that the complex AI calculations were performed correctly and without any sneaky interference. Furthermore, Flare Time Series Oracle (FTSO), Flare Data Connector (FDC), and Flare’s built-in data tools are perfectly crafted to ensure data history and trustworthy verification for AI systems.

Boasting over 460 eager participants across four different tracks, the hackathon sparked the creation of verified AI solutions in diverse areas, from collaborative learning approaches like Consensus Learning to the exciting intersection of AI and decentralized finance (DeFI), dubbed DeFAI. Every project utilized Confidential Space instances with virtual Trusted Platform Module (vTPM) attestations, running on the robust AMD Secure Encrypted Virtualization (SEV) platform. Think of these attestations as cryptographic fingerprints, providing solid proof that computations happened inside a secure and tamper-proof space.

Rene Kolga, Product Manager at Google Cloud, shared his enthusiasm:

”It’s inspiring to witness developers harnessing the power of our Confidential Space secure enclaves to craft trustworthy AI solutions. Projects like 2DeFi and the unveiling of the Flare AI Kit are truly exciting milestones, paving the way for a future where AI and blockchain work hand-in-hand to guarantee data integrity and reliable computation.”

2DeFi: Seamless TradFi to DeFi Transition with AI

One standout project, 2DeFi, was developed by two talented University of Waterloo students, Alex and Hitarth. Their innovative idea not only secured second place in the RAG Knowledge track but also clinched the top spot in the AI x DeFi (DeFAI) track.

2DeFi ingeniously employed Google Gemini AI to bridge the gap between traditional finance (TradFi) customers and the world of DeFi on Flare. Imagine this: users could simply upload screenshots of their Robinhood portfolios, and Gemini 2.0 would analyze them to gauge their risk tolerance. Then, based on this AI-generated risk profile, 2DeFi would automatically map out a suitable DeFi strategy, including staking and providing liquidity on the Flare network.

By integrating embedded wallets, 2DeFi eliminated the usual complexities of getting started with Web3. Users could set up their Flare wallets simply using their Google login, dramatically improving the user experience and making onboarding a breeze.

Flare AI Kit: Empowering Verifiable AI Agents on Blockchain

Excitingly, the hackathon also marked the official launch of the Flare AI Kit, an open-source toolkit designed to help developers build verified AI agents on Flare, leveraging Google Cloud Confidential Space. The Flare AI Kit was put to the test in various hackathon projects, showcasing its potential as a foundational platform for AI-powered blockchain applications. It provides developers with robust security and trustworthy, verifiable AI execution environments, simplifying the integration of AI while preserving verifiability, data origin tracking, and AI safety standards.

Born from the Flare x Google Cloud Hackathon, the Flare AI Kit is poised to evolve into a powerful, production-ready platform for verifiable AI applications, fueled by ongoing exploration at the crossroads of blockchain and artificial intelligence by developers and academics. The ultimate goal is to transform the Flare AI Kit into a ready-to-use solution, accessible on the Google Cloud Marketplace. This will empower developers and businesses alike to adopt it as a standalone, production-grade tool, continuously enriched with enhanced features, improved usability, and broader integrations.

The Flare AI Kit aims to revolutionize how verifiable AI applications are developed, deployed, and trusted worldwide, seamlessly blending Confidential Space TEEs, AI, and the Flare blockchain. Ready to start building on Flare? Dive into the Flare Developer Hub for comprehensive resources and documentation to get you going.