🚀 Building AI Oracle for Social Media Verification - Looking for Collaborators

Developers
1

The Problem: Influencer marketing loses $1.3B+ yearly to fake followers and bot engagement. No trustless verification exists.

The Solution: AI-powered oracle (like on oraichain) that verifies social media milestones (follower growth, engagement authenticity, bot detection) and triggers smart contract payments automatically.

Tech Stack:

  • PyTorch (bot detection, growth pattern analysis, authenticity scoring)

  • Rust (high-performance oracle API)

  • ICP Smart Contracts (escrow & payment logic)

  • Twitter/Instagram APIs

What We’re Building: Multi-model ML system that analyzes followers for authenticity → returns verified results to blockchain → releases payment if conditions met.

Looking For:

  • ICP/Motoko developers (smart contract escrow logic)

  • ML engineers (fraud detection, graph analysis)

  • Backend developers (API scraping, rate limit handling)

Your Benefits: :white_check_mark: Real-world AI + Web3 project for portfolio :white_check_mark: Solve actual $1B+ problem :white_check_mark: Rare skill combination (ML + blockchain) :white_check_mark: Stand out to employers in both spaces :white_check_mark: Open source contribution

My Background: Senior dev with AI/ML expertise + Rust + Solana experience. Now exploring ICP ecosystem.

Ready to Build? DM me or comment below. Let’s create infrastructure that makes influencer marketing trustless.

3 Likes
2

High-Level Architecture (30-Day Build)

1. Data Collection Layer

  • X.com API - Fetch follower data, engagement metrics, account details

  • Custom Scraper - Backup/supplementary data when API limited

  • Store raw data → preprocessed datasets

2. ML Development (Google Colab)

  • Research phase - Experiment with models, feature engineering

  • Models to build:

    • Bot detection classifier (follower account analysis)

    • Engagement authenticity scorer (NLP on comments/interactions)

    • Growth pattern analyzer (time-series anomaly detection)

    • Network graph analyzer (detect bot farm clusters)

  • Train, validate, export optimized models (ONNX format)

3. Inference Layer (Rust API)

  • Load trained models (ONNX Runtime for fast inference)

  • Accept requests: {username, milestone_type, target_count}

  • Orchestrate: scrape → run models → aggregate scores → return verdict

  • Caching layer for repeat queries

  • Rate limiting and error handling

4. Oracle Bridge (ICP Smart Contract)

  • Escrow logic: hold funds until conditions met

  • Call Rust API for verification

  • Receive ML results → execute payment or reject

  • Emit events for transparency

5. Optional: Claude/LLM Integration

  • Use Claude API for explainability: “Why was this flagged?”

  • Generate human-readable reports from ML scores

  • Content quality analysis (post relevance to niche)

6. Deployment

  • Google Colab - Model training/research only

  • Cloud GPU (Nvidia/RunPod/Lambda) - If heavy inference needed

  • Standard VPS - Likely sufficient for ONNX inference (CPU-based)

  • ICP Canister - Smart contract deployment

30-Day Timeline:

Week 1: Data collection + scraper + basic bot detection model Week 2: Train/optimize models in Colab, export to ONNX Week 3: Build Rust API, integrate models, test inference Week 4: ICP smart contract, oracle bridge, end-to-end testing

Clarification: You don’t need “Claude server” for training. Claude API is for inference/reasoning tasks only. For GPU training, use Colab (free tier) or rent GPUs if needed.

2 Likes