​
Overview

Kyne creates a decentralized AI marketplace by combining SUI blockchain’s programmable transactions with Walrus’s efficient storage network. This enables:
  • Secure AI model and dataset trading
  • Verifiable compute for AI training
  • AI-to-earn incentives for data contributors
  • Decentralized storage with 4-5x efficiency vs competitors

​
Core Components

SUI Blockchain Layer

  • • Handles all marketplace transactions
  • • Manages economic incentives and rewards
  • • Verifies compute results and proofs
  • • Governs protocol parameters

Walrus Storage Layer

  • • Stores AI models and datasets
  • • Uses Red Stuff encoding for 4-5x efficiency
  • • Ensures Byzantine fault tolerance
  • • Provides verifiable storage proofs

​
How It Works

​
Key Features

​
AI Data Marketplace

For Data Contributors

  • • Earn tokens for quality data contributions
  • • Verifiable usage tracking
  • • Revenue share from AI models
  • • Stake to earn storage rewards

For AI Developers

  • • Access curated datasets
  • • Deploy models with verifiable compute
  • • Pay only for storage used
  • • Built-in monetization

​
Tokenomics & Incentives

The protocol uses SUI’s native capabilities for:
  • Data contribution rewards
  • Storage node staking
  • Compute verification
  • Marketplace fees
  • Protocol governance

​
Storage Architecture

Walrus provides efficient decentralized storage through:
  1. Red Stuff Encoding
    • 4-5x more efficient than full replication
    • Byzantine fault tolerant
    • Fast recovery from node failures
  2. Storage Proofs
    • Verifiable on SUI blockchain
    • Automatic challenge/response
    • Stake slashing for violations
  3. Economic Security
    • Storage nodes stake tokens
    • Rewards for reliable storage
    • Penalties for data loss

​
For Developers

​
Quick Start

from kyne.sdk import KyneClient

# Initialize client
client = KyneClient(
    sui_endpoint="https://sui-rpc.example.com",
    walrus_endpoint="https://walrus.example.com"
)

# Upload dataset and create marketplace listing
dataset_id = await client.upload_dataset(
    path="./my_dataset",
    metadata={
        "name": "Training Data v1",
        "price": 1000, # in SUI tokens
        "reward_share": 0.1 # 10% to contributors
    }
)

# Deploy AI model
model_id = await client.deploy_model(
    model_path="./my_model",
    input_dataset=dataset_id,
    compute_config={
        "gpu": "t4",
        "batch_size": 32
    }
)

​
Storage Integration

# Store data with Walrus
blob_id = await client.store_blob(
    data=my_data,
    config={
        "min_nodes": 7,     # 2f+1 for Byzantine fault tolerance
        "epochs": 12,       # Storage duration
        "redundancy": 2     # Red Stuff encoding level
    }
)

# Verify storage proof on SUI
proof = await client.verify_storage(blob_id)

​
Security Model

​
Data Protection

​
Access Control

On-Chain

  • • SUI smart contract permissions
  • • Token-gated access
  • • Verifiable compute results
  • • Automated payments

Storage Layer

  • • Encrypted data storage
  • • Byzantine fault tolerance
  • • Storage proofs
  • • Node stake requirements

​
Next Up

Model Pipeline

Learn how to train and deploy models using Kyne’s distributed compute network.

Search Infrastructure

Explore Kyne’s distributed search and indexing capabilities for AI data.
Kyne’s infrastructure enables efficient model training and intelligent data discovery through distributed compute and search capabilities.