Satya Protocol

Excited to share my latest hackathon project: Satya, a decentralized AI model marketplace built on the Sui stack. For a while now, I've been researching bridging AI and Web3, and I have noticed some gaps in today’s AI ecosystem. Centralized platforms still struggle with trust: creators face piracy and unfair payouts, while buyers gamble on fake, biased, or underperforming models. So I built Satya to flip that dynamic, a trustless, cryptographically verifiable marketplace where creators keep control and buyers know exactly what they’re getting. Satya brings verifiable trust and strong IP protection to AI models. Using TEEs via Nautilus, every model gets tamper-proof attestations for authenticity, performance, and bias, all visible on-chain. No more blind buys. For creators, Seal (secrets management on @WalrusProtocol) enables encryption so models stay encrypted even during evaluations, preventing theft or reverse-engineering. Sui smart contracts automate payouts, fees, and revenue splits with sub-second finality, and dynamic NFTs let model metadata evolve as new verifications come in. On the storage side, Walrus provides decentralized, redundant, blob-based hosting for encrypted models, giving creators reliability and buyers verifiable access without centralized points of failure. I built this as a prototype for the Walrus Foundation Haulout Hackathon, using Sui’s object-centric model and parallel execution to power high-throughput AI asset trading. Satya isn’t just another marketplace, it’s a verification-first ecosystem that treats AI models as secure, tradable digital assets. Huge thanks to @Mysten_Labs for the powerful tooling.

@minting_ruru @SatyaProtocol @SuiNetwork @EmanAbio @theharrisonkim This actually makes a lot of sense. I love how the whole process stays secure from upload to download. Satya really figured out a clean way to protect creators while still making it easy for buyers.

The model download process on @SatyaProtocol: TEE verification → Onchain Verification → Model Purchase → buyer requests decryption → Key manager releases session key → Walrus blob decrypted → Model downloaded by buyer When a creator uploads a model on Satya, it first goes into a "pending". At that point, the model is already encrypted with Seal and stored on @WalrusProtocol, but nothing is live on the marketplace yet. It just sits in the user dashboard while the platform queues it for verification. All we have at this stage is the encrypted blob, the on-chain metadata, and the listing skeleton. From there, the model gets pushed into a TEE for verification. The enclave runs the model, checks that it has not been modified, benchmarks it, and then returns a hardware attestation that proves the environment was real and secure. Once that attestation comes back and gets written to the listing on Sui, the model automatically moves from Pending to Verified. At that moment it becomes visible and can be purchased like any other item on the marketplace. When someone purchases the model, the purchase goes through a Sui smart contract that handles payment and access rights. After the purchase is confirmed, the buyer can request decryption, and the SEAL key manager checks their proof of ownership before issuing a temporary session key. That key is used to decrypt the file stored on Walrus, and the buyer receives a secure download link.

Over the past couple of weeks, I've been building an AI Model Marketplace powered by entire Sui stack, designed for a future where data, compute, and intelligence flow securely across open networks. My work focuses on solving one of the biggest problems in AI today, how to enable developers, companies, and creators to exchange models without giving up ownership or exposing sensitive information. By combining verifiable storage - @WalrusProtocol, with hardware-backed confidentiality, using the Nautilus framework on @SuiNetwork , the marketplace makes trust an engineering property, not a negotiation. At the core of this project is the belief that AI shouldn’t be gated behind closed platforms. TEEs allow models to run in sealed environments where inputs, outputs, and weights remain private, while @WalrusProtocol ensures data availability, verifiability, and censorship resistance, leveraging Seal. Together, they create a system where anyone can deploy a model, sell access, prove correct execution, and earn, without ever losing control of their intellectual property. This marketplace is built for the next generation of AI builders, researchers needing privacy guarantees, startups shipping intelligent agents, enterprises protecting proprietary knowledge, and on-chain apps requiring verifiable computation. The goal is simple, unlock a permissionless AI economy where value flows transparently, securely, and globally. cc: @theharrisonkim, @RJ_Simmonds.

the model upload process on @SatyaProtocol: data entry -> walrus upload + seal encryption -> pending in user dashboard

the model upload process on @SatyaProtocol: data entry -> walrus upload + seal encryption -> pending in user dashboard