AI2Work Case Study — Atlas Nexus

Challenge

The Problem

Freelancers lose 5-20% in platform fees on traditional marketplaces. Payment disputes take days. Cross-border payments add 3-5% forex and 2-3 business days. Existing Web3 freelance platforms are desktop-only and require MetaMask — inaccessible to the 70% of global users on mobile.

Approach

The Solution

AI2Work is a mobile-first Web3 freelancer marketplace on Celo — carbon-negative, mobile-optimized L1. Freelancers post bounties in cUSD, CELO, or USDC. Funds held in auditable smart contract escrow. 2% protocol fee (vs 5-20%). Settlement: instant on-chain. No KYC. No bank. Works on any Android device.

Stack

Architecture

Smart Contracts

Solidity · Foundry · OpenZeppelin v5

• ClaudelanceCore.sol — escrow, dispute resolution, stake settlement
• ReentrancyGuard + Ownable2Step + Pausable
• 20+ mutating functions, tested with forge test
• Deployed on Celo Mainnet (chain 42220), verified on Celoscan

Web PWA

Next.js 15 · TypeScript · Tailwind · WalletConnect v2

• Server-side rendering with App Router
• PWA with offline support, installable on mobile
• Hosted on Render.com, CI/CD via GitHub Actions

Native Android

Kotlin · Capacitor · Android Keystore

• Custom Capacitor plugin for ECDSA signing
• Android Keystore for private key storage (hardware-backed)
• Keccak256 implemented from scratch in Kotlin
• EIP-1559 transaction signing + RLP encoding
• WebView bridge injecting window.ethereum

SDK & Types

npm packages · typed client

• @yeheskieltame/claudelance-sdk — typed contract client
• @yeheskieltame/claudelance-types — shared ABI + TypeScript types
• Published on npm + GitHub Packages

Deep Dive

Native ECDSA on Android

The core innovation is a custom Capacitor plugin that implements Ethereum transaction signing natively in Kotlin — no Web3 wallet required. This eliminates the MetaMask dependency that blocks mobile Web3 adoption.

Key decisions

• secp256k1 via BouncyCastle (SpongyCastle on Android)
• Keccak256 implemented manually — no extra dependency
• RLP encoding for transaction serialization
• Android Keystore for key generation (hardware-backed when available)
• Bridge.js injected into WebView — dapps see a standard window.ethereum

// EthWalletHelper.kt — Keccak256 core
fun keccak256(data: ByteArray): ByteArray {
    val digest = Keccak(256)
    return digest.digest(data)
}

// EIP-1559 transaction signing
fun signTransaction(tx: EthTransaction, key: PrivateKey): ByteArray {
    val encoded = rlpEncode(tx.toList())
    val hash = keccak256(encoded)
    return signECDSA(hash, key) // secp256k1 deterministic
}
      

Metrics

Results & Impact

Gas cost per bounty~$0.01 (Celo)

Protocol fee2%

APK size4.2 MB

Android API24+ (99.1% devices)

Transaction signingNo external wallet needed

Retro

Lessons Learned

1

Capacitor plugins are the right abstraction for Web3 mobile — they bridge the gap between web UX and native crypto without forcing users into a separate wallet app.

2

Celo's sub-cent gas fees make on-chain freelancing viable. On Ethereum mainnet, a single bounty posting would cost $5-20 in gas. On Celo: $0.01.

3

The hardest part of Web3 mobile is key management. Android Keystore + hardware backing is the only acceptable solution — no plaintext keys in SharedPreferences.

4

The Celo ecosystem's focus on mobile-first and real-world assets (cUSD, USDC) aligns perfectly with freelancer payments — stablecoins are non-negotiable for professional use.

Want something like this?

I build full-stack Web3 products — smart contracts, mobile apps, agent orchestration.

Get in touch