Decentralized
AI Inference

A trustless protocol for AI model evaluation through cryptographic commit-reveal schemes and peer-reviewed consensus mechanisms.

<400ms
Commit-Reveal Latency
3validators
Peer Review Threshold
99.7%
Consensus Accuracy
<0.1%
Slashing Rate

Get Started

Choose your path to start using the RIFT protocol

Contest Creators

DAOs, Enterprises, AI Labs

RECOMMENDEDdApp UI

Web Interface

  • • Connect wallet (Phantom/Solflare)
  • • Fill form: duration, rewards, parameters
  • • Fund reward pool
  • • One-click create
Open Dashboard
INTEGRATIONSDK

TypeScript SDK

npm install @riftedai/sdk
const rift = new RiftClient(connection, wallet);
await rift.createContest({ duration: 7d, reward: 50000 });
SDK Docs
POWER USERSCLI

Command Line

npm install -g @riftedai/cli
rift create-contest --duration 7d --reward 50000
rift fund-pool 1 --amount 10000
CLI Reference

Participants

AI Developers, ML Engineers, Researchers

MOST ACCESSIBLEdApp UI

Web Interface

  • • Browse active contests
  • • Join with stake
  • • Upload submission via UI
  • • Review others & claim rewards
Start Competing
RECOMMENDED FOR DEVSCLI

Command Line

rift join 1 --stake 100
rift submit --result output.json --model gpt4
rift reveal 1
rift claim 1
CLI Reference
AUTOMATIONSDK

Pipeline Integration

// Auto-submit after ML inference
const output = await model.inference(prompt);
const attestation = await tee.attest(output);
await rift.commitSubmission(1, { output, attestation });
Integration Guide

The RIFT Protocol

A permissionless infrastructure for evaluating AI model outputs through game-theoretic incentives. Participants stake tokens, submit inference results, and earn rewards based on peer-validated quality scores.

Byzantine fault tolerant consensus
TEE-verified computation attestations
Sybil-resistant reputation system
Zero-knowledge commit-reveal scheme
protocol_flow.rs
pub fn submit_inference(
    ctx: Context<SubmitInference>,
    commitment: [u8; 32],
) -> Result<()> {
    let submission = &mut ctx.accounts.submission;
    submission.commitment = commitment;
    submission.timestamp = Clock::get()?.unix_timestamp;
    
    transfer_tokens(
        ctx.accounts.participant,
        ctx.accounts.vault,
        ctx.accounts.stake_amount,
    )?;
    
    emit!(InferenceCommitted {
        participant: ctx.accounts.participant.key(),
        commitment,
    });
    
    Ok(())
}

Protocol Architecture

A modular stack designed for trustless AI evaluation at scale

Layer 4
Application Layer
Contest InterfaceSubmission PortalReview Dashboard
Layer 3
Protocol Layer
Commit-Reveal EnginePeer Review ConsensusReward Distribution
Layer 2
Verification Layer
TEE AttestationZK ProofsSignature Validation
Layer 1
Settlement Layer
Solana RuntimeSPL TokenPDA State

Protocol Flow

A four-phase mechanism ensuring fair and verifiable AI evaluation

01

Commit

Participants submit cryptographic commitments of their inference results.

keccak256(result || attestation || salt)
02

Reveal

After the commit window closes, participants reveal their actual results.

verify(commitment, result, salt) → bool
03

Review

Peer reviewers evaluate submissions with reputation-weighted scoring.

score = Σ(weight_i × review_i) / Σ(weight_i)
04

Settle

Rewards distributed proportionally. Malicious actors slashed.

reward = (score / total_score) × pool

Deployed Contracts

Verified and deployed on Solana Devnet

SPL token with controlled minting
RIFT Token Program
5kCioito1CpAxEuqaMAQsyrZ1gt5tQQ8qnKzG3V816gb
Core protocol logic and state
Contest Protocol
35UXUBPvYUG5zqnMRfnGduD5WBnQ6qWadW7m3gaaKSD1

Build with RIFT

Integrate decentralized AI evaluation into your applications.