# Kweli: AI + ZK Deepfake Detection Platform
Kweli is a youth-centered, privacy-preserving platform that combines AI-based deepfake detection with Zero-Knowledge (ZK) cryptography to verify digital media authenticity. Built for communities, schools, and mobile-first users in Africa, Kweli empowers users to fight misinformation with truth, proof, and transparency.
⚖️ Kweli is a work-in-progress to empower African institutions and civic actors with privacy-first truth verification tools and powered by Zero-Knowledge cryptography.
📸 From verifying deepfakes to securing media evidence, Kweli is building tools for digital justice in the age of disinformation.
🔐 Support the mission: Sponsor our work on GitHub
---
## 🌍 Project Goals
- **Detect and flag deepfakes** using an efficient AI model (EfficientNet-lite).
- **Verify authenticity** of media using ZK proofs stored on Starknet blockchain.
- **Educate youth and educators** with culturally-relevant digital literacy toolkits.
- **Enable trust at scale** with decentralized infrastructure and community testing.
---
## 🧠 Tech Stack
- **Frontend**: Expo + React Native
- **Backend**: Flask (Python), IPFS
- **AI Model**: EfficientNet-lite (TensorFlow/Keras)
- **Blockchain**: Starknet (Cairo 2, Sierra, Starknet.py)
- **ZK Layer**: Cairo ZK circuits for proof generation and validation
- **Dev Tools**: EAS, VSCode, GitHub Actions
---
## 📦 Key Features
- Image and video upload with live fake/real prediction
- ZK-based verification with hash commitments stored on-chain
- A/B tested UI for accessibility
- Offline caching and async uploads for low-connectivity areas
- Real-time logging of false positives and negatives
---
## 📅 Milestone-Driven Development
The project is executed over 12 months with quarterly milestones:
- **Q1**: Pilot rollout, onboarding, user testing
- **Q2**: AI model optimization, ZK module, blockchain stress tests
- **Q3**: Curriculum co-development, community testing
- **Q4**: Policy advocacy, na …