|
From July 29 th, 2026, Cakradana will transition to new domains as 🌐 Website: cakradana.faizath.com (formerly cakradana.org) |
Empowering Clean & Transparent Elections with AI-Powered AML Detection
- 🎯 About Cakradana
- 🏆 TrackAML 2.0 Achievement
- 🎬 Project Showcase Video
- ❌ The Problem
- ✅ How Cakradana Solves It
- 📸 Screenshots
- 📚 Repository Overview
- 🛠️ Tech Stack
- 🚀 Quick Start
- 🤝 Contributing
- 📞 Contact
Cakradana is an artificial intelligence (AI) system specifically designed to enhance transparency and detect risks in election financing in Indonesia. This platform uses advanced technology to analyze, process, and visualize political donation data in real-time.
The name "Cakradana" is derived from Sanskrit meaning "wheel of giving" - symbolizing the cycle of transparency in a healthy democratic system.
Cakradana in Action - AI-Powered Election Finance Transparency
TrackAML 2.0 Hackathon 2025 is a national innovation competition organized by PPATK (Indonesia's Financial Intelligence Unit) as part of the Gerakan Nasional APUPPT & PPSPM — Indonesia's national movement against money laundering and terrorism financing. The hackathon challenges participants to build AI- and technology-driven solutions across three themes: Environmental & Financial Sustainability (e.g. illegal carbon trading), Organized Crime & Transparency (corruption, fraud, suspicious transaction monitoring), and Transnational Crime Prevention (human trafficking, narcotics, terrorism financing). It runs two competition tracks — a Public/Private Sector track for government agencies and financial institutions, and an Open Sector track for academics, researchers, startups, and students — culminating in a final competition and awarding ceremony.
Cakradana was submitted for TrackAML 2.0 Hackathon hosted by PPATK (Pusat Pelaporan dan Analisis Transaksi Keuangan), Indonesia's Financial Transaction Reports and Analysis Center and achieved 🥉 3rd Winner in the Open Sector category.
💰 Prize Awarded: IDR 20,000,000 (~USD 1,200) for our innovative AI-powered solution for election finance transparency.
This recognition validates our commitment to building innovative solutions for financial transparency and anti-money laundering in Indonesia's electoral system.
The award was directly presented by the Head of PPATK, Dr. Ivan Yustiavandana, S.H., LL.M., to the Cakradana team at the TrackAML 2.0 International Hackathon awarding ceremony. (Official PPATK bio)
Click the image above to watch the Cakradana project showcase video.
Political campaign financing in Indonesia is highly exposed to money laundering — smurfing, donations disguised as "self-funded" contributions, proxy accounts, and unclear donor identities that regulators cannot trace back to their source.
According to Indonesia Corruption Watch (2023), corruption cases tied to public-fund misuse rose 15% year-over-year, with potential state losses reaching into the trillions of rupiah. Existing AML systems were built to monitor banking transactions, not political donation patterns — they can't map donor↔candidate relationships or cross-check campaign reports against real transaction data.
The scale of the blind spot:
- 27.8 million transaction reports reached PPATK in 2022 — yet only 0.3% were categorized as Suspicious Transaction Reports (STR).
- 90–95% of AML system alerts are false positives, creating heavy compliance workload with low efficiency.
- The average time to detect and report money laundering activity is 151 days.
- Rp195 billion in foreign funds were traced into the accounts of 21 political parties ahead of the 2024 election — despite foreign campaign funding being illegal.
- PPATK flagged suspicious transactions tied to 100 legislative candidates totaling Rp51.4 trillion, with no visibility into donor identity or fund routing.
Root cause: KPU, PPATK, and political parties each hold a piece of the data, but no integrated system connects the dots, reads the transaction patterns, or surfaces hidden money-laundering practices — so everything can look legitimate on paper.
Cakradana is an AI-powered monitoring and risk-detection system built to close this blind spot end-to-end:
- 🎯 Real-Time Donation Risk Scoring — a LightGBM machine learning model scores every donation 0–100 in real time, achieving over 85% accuracy in flagging potentially suspicious activity such as smurfing or layered donations.
- 📄 Multi-Source Data Integration — digital donation forms and paper-based forms are unified into structured data via OCR (Tesseract.js), so no submission channel is a blind spot.
- 🧠 NLP-Powered Standardization — a self-hosted, open-source LLM (DeepSeek) cleans and standardizes the unstructured text extracted by OCR into analyzable data.
- 🕸️ Interactive Network Visualization — a graph interface maps relationships between donors (individuals, companies, parties) and recipients, making complex donation networks easy to explore and audit.
- 🛡️ Candidate Verification Dashboard — candidates and recipients can log in to confirm or dispute flagged transactions, adding a human verification layer on top of automated detection.
- 📋 Standardized Suspicious Activity Reports — findings are exported in PPATK/FATF-compliant SAR format, ready for regulatory action.
- 🔒 On-Premise Deployment — runs on-premise to meet government-grade data security requirements.
Measured impact:
| Metric | Before | With Cakradana |
|---|---|---|
| Suspicious transaction detection rate | 0.3% | ~5% (17x improvement) |
| Reporting turnaround time | 151 days | Under 14 days |
| False positive rate | 90–95% | ~30–45% |
Beyond the numbers, Cakradana increases public transparency through interactive donation network visualizations and strengthens candidate accountability with a complete digital audit trail.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Cakradana consists of three main repositories, each serving a specific purpose in the ecosystem:
AI/ML Engine for Risk Detection
- Purpose: Core machine learning system for detecting risky political donations
- Features:
- LightGBM-based risk scoring model
- Feature engineering (temporal, network, behavioral)
- Rule-based compliance checks
- FastAPI service for real-time predictions
- Tech Stack: Python, FastAPI, LightGBM, scikit-learn, pandas
- Model Performance: ROC-AUC 0.84, optimized for recall of risky transactions
Backend API & Data Processing
- Purpose: Central API service handling data ingestion and processing
- Features:
- Multi-format data input (digital forms, OCR, web scraping)
- Authentication & user management
- Network analysis and entity tracking
- RESTful API with OpenAPI documentation
- Tech Stack: Node.js, Express.js, MongoDB, Tesseract.js, JWT
- Live API: https://cakradana-api.faizath.com
Frontend Web Application
- Purpose: User interface for donation monitoring and analytics
- Features:
- Real-time dashboard with interactive visualizations
- Network graph visualization
- Risk analysis reports
- User authentication and role-based access
- Tech Stack: Next.js, React, TypeScript, Tailwind CSS, Chart.js
- Live Website: https://cakradana.faizath.com
- Python 3.10+ - Core programming language
- FastAPI - High-performance API framework
- LightGBM - Gradient boosting for risk detection
- scikit-learn - Machine learning utilities
- pandas - Data manipulation and analysis
- joblib - Model serialization
- Node.js 18+ - Runtime environment
- Express.js - Web application framework
- MongoDB - NoSQL database with Mongoose ODM
- Tesseract.js - OCR for document processing
- JWT - Stateless authentication
- Multer - File upload handling
- Nodemailer - Email service integration
- Next.js 15.4.2 - React framework with App Router
- React 18.3.1 - UI library
- TypeScript 5.0 - Type safety
- Tailwind CSS 3.4.17 - Utility-first CSS framework
- Chart.js - Interactive data visualization
- D3.js - Advanced data visualization
- GSAP - Smooth animations
- Docker - Containerization
- Cloudflare R2 - Object storage
- OpenRouter API - LLM integration
- GitHub - Version control and CI/CD
- Website: https://cakradana.faizath.com
- API Documentation: https://cakradana-api.faizath.com
# Clone all repositories
git clone https://github.com/cakradana-app/cakradana-ai.git
git clone https://github.com/cakradana-app/cakradana-api.git
git clone https://github.com/cakradana-app/cakradana-web.git
# Run AI service
cd cakradana-ai
pip install -r requirements.txt
uvicorn app_fastapi:app --host 0.0.0.0 --port 8000
# Run API service
cd ../cakradana-api
npm install
npm run dev
# Run Web application
cd ../cakradana-web
npm install
npm run dev# Test risk prediction
curl -X POST "http://localhost:8000/predict" \
-H "Content-Type: application/json" \
-d '{
"instances": [{
"sender_type": "individual",
"receiver_type": "political-party",
"amount": 2000000
}]
}'Get in touch with the Cakradana team
Made with ❤️ by the Cakradana Team
Empowering democracy through transparency and technology
© 2025 Cakradana. Building a more transparent democracy through technology.