Synth Studio 🧪

Privacy-first synthetic data generation for healthcare and fintech

---

⚡ Quick Install

# Clone
git clone https://github.com/Urz1/synthetic-data-studio.git && cd synthetic-data-studio

# Backend
cd backend && cp .env.example .env
pip install -r requirements.txt && alembic upgrade head
uvicorn app.main:app --reload

# Frontend (new terminal)
cd frontend && cp .env.local.example .env.local
pnpm install && pnpm dev

Frontend: http://localhost:3000 | API Docs: http://localhost:8000/docs

📖 Full setup guide: LOCAL_DEVELOPMENT.md

---

🎯 What It Does

Generate high-quality synthetic data with differential privacy guarantees. Built for regulated industries:

Industry	Use Case
🏥 Healthcare (HIPAA)	Synthetic EHR, FHIR, patient records
🏦 Fintech (SOC-2/GDPR)	Transaction data, fraud testing
🤖 ML Teams	Privacy-safe training datasets
🏢 Enterprise	Cross-department data sharing

---

✨ Key Features

Generation Methods

Method	Description	Best For
Schema-Based	Define columns → generate data (no source dataset needed)	Testing, prototyping
Dataset-Based ML	Train on real data → generate synthetic	Production quality
LLM-Powered Seed	AI generates realistic seed data → statistical expansion	Domain-specific realism

ML Generators

• CTGAN - Conditional Tabular GAN (mixed numeric + categorical)
• TVAE - Tabular Variational Autoencoder (high-cardinality categorical)
• GaussianCopula - Statistical copulas (fast, correlation-preserving)

Privacy & Compliance

• Differential Privacy - Configurable ε/δ with RDP accounting
• PII/PHI Detection - Automatic sensitive column identification
• Compliance Reports - HIPAA, GDPR, SOC-2 ready documentation
• Audit Logs - Immutable activity tracking

AI-Powered Features

• Chat Assistant - Natural language data generation guidance
• Enhanced PII Detection - LLM-powered sensitivity analysis
• Compliance Writer - Auto-generate compliance documentation

Quality Evaluation

• Statistical Similarity - Distribution matching, K-S tests
• ML Utility - Train/test accuracy preservation
• Privacy Risk - Membership inference, re-identification risk

---

📋 Prerequisites

Requirement	Version
Python	3.9+
Node.js	18+
PostgreSQL	13+
Redis	7+ (local Docker by default; set REDIS_URL for managed)

Environment Variables:

# Backend (.env)
DATABASE_URL=postgresql://user:pass@localhost/synthstudio
SECRET_KEY=your-jwt-secret
AWS_S3_BUCKET=your-bucket  # optional
REDIS_URL=redis://localhost:6379/0  # default local container; use rediss:// for hosted

# Frontend (.env.local)
NEXT_PUBLIC_API_URL=http://localhost:8000
BETTER_AUTH_SECRET=your-auth-secret

---

🔧 Usage

Generate from Schema (No Dataset Needed)

curl -X POST "http://localhost:8000/generators/schema" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "columns": {
      "name": {"type": "string", "faker": "name"},
      "age": {"type": "integer", "min": 18, "max": 80},
      "email": {"type": "string", "faker": "email"},
      "balance": {"type": "number", "min": 0, "max": 50000}
    }
  }'

Generate from Dataset (ML-Based)

# Upload dataset
curl -X POST "http://localhost:8000/datasets/upload" \
  -H "Authorization: Bearer $TOKEN" \
  -F "file=@data.csv"

# Generate synthetic data with DP
curl -X POST "http://localhost:8000/generators/dataset/{dataset_id}/generate" \
  -H "Authorization: Bearer $TOKEN" \
  -d '{
    "generator_type": "ctgan",
    "num_rows": 10000,
    "epochs": 300,
    "differential_privacy": {"enabled": true, "epsilon": 1.0, "delta": 1e-5}
  }'

Python SDK Example

import requests

# Login
session = requests.Session()
session.post("http://localhost:8000/auth/login", json={
    "email": "user@example.com", "password": "secret"
})

# Schema-based generation
synth_data = session.post("/generators/schema?num_rows=1000", json={
    "columns": {
        "patient_id": {"type": "string", "pattern": "PAT-[0-9]{6}"},
        "diagnosis": {"type": "category", "values": ["A01", "B12", "C34"]},
        "visit_date": {"type": "date", "min": "2024-01-01", "max": "2024-12-31"}
    }
}).json()

---

🧪 Testing

# Backend tests with coverage
cd backend && pytest tests/ -v --cov=app

# Frontend tests
cd frontend && pnpm test

# E2E tests
cd frontend && pnpm test:e2e

---

📁 Project Structure

synth-studio/
├── backend/                  # FastAPI API server
│   ├── app/
│   │   ├── auth/            # Authentication (JWT, OAuth, 2FA)
│   │   ├── datasets/        # Dataset upload, profiling
│   │   ├── generators/      # Schema + ML generation
│   │   ├── evaluations/     # Quality metrics
│   │   ├── services/
│   │   │   ├── synthesis/   # CTGAN, TVAE, Copula
│   │   │   ├── llm/         # AI chat, PII detection
│   │   │   └── privacy/     # DP accounting
│   │   ├── compliance/      # HIPAA/GDPR reports
│   │   └── audit/           # Activity logging
│   └── tests/
├── frontend/                 # Next.js 16 web app
│   ├── app/
│   │   ├── dashboard/       # Overview & metrics
│   │   ├── datasets/        # Upload & profile
│   │   ├── generators/      # Create & manage
│   │   ├── evaluations/     # Quality reports
│   │   ├── synthetic-datasets/  # Generated data
│   │   ├── compliance/      # Compliance center
│   │   └── assistant/       # AI chat
│   └── components/
└── docs/                     # Docusaurus docs

---

📚 Documentation

Resource	Description
Docs Site	Full documentation
Getting Started	Installation & quickstart
User Guide	Feature walkthroughs
API Reference	OpenAPI/Swagger
Examples	Code samples & Postman

---

🤝 Contributing

1. Fork & clone
2. Create feature branch (git checkout -b feature/amazing)
3. Add tests & make changes
4. Run tests (pytest / pnpm test)
5. Submit PR

See CONTRIBUTING.md for guidelines.

---

🔒 Security

Report vulnerabilities privately: halisadam391@gmail.com or see SECURITY.md.

---

📄 License

MIT © 2025 Sadam Husen

---

📬 Contact

Sadam Husen @Urz1 halisadam391@gmail.com

LinkedIn • GitHub

---

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│               Frontend (Next.js 16)                 │
│  Dashboard • Datasets • Generators • Evaluations   │
│  Compliance • Audit • Billing • AI Assistant       │
└────────────────────────┬────────────────────────────┘
                         │ REST API (JWT + OAuth)
┌────────────────────────▼────────────────────────────┐
│                Backend (FastAPI)                    │
│  ┌──────────┐ ┌──────────┐ ┌──────────┐            │
│  │   Auth   │ │ Datasets │ │Generators│            │
│  │JWT/OAuth │ │Profiling │ │Schema/ML │            │
│  └──────────┘ └──────────┘ └──────────┘            │
│  ┌──────────┐ ┌──────────┐ ┌──────────┐            │
│  │   LLM    │ │Evaluation│ │Compliance│            │
│  │Chat/PII  │ │Quality   │ │Reports   │            │
│  └──────────┘ └──────────┘ └──────────┘            │
└───────┬───────────────┬────────────────┬────────────┘
        ▼               ▼                ▼
   PostgreSQL        Redis          AWS S3
   (metadata)     (queue/cache)    (files)
                        │
                        ▼
              Celery Workers
        (generation, evaluation, exports)

Tech Stack:

• Frontend: Next.js 16, React 19, TypeScript 5, Tailwind, shadcn/ui
• Backend: FastAPI, SQLAlchemy 2, Celery, SDV
• ML/Privacy: CTGAN, TVAE, Opacus (DP), RDP accounting
• LLM: OpenAI/Anthropic (chat, PII detection, compliance)
• Infra: Vercel, Railway/AWS, Neon/Supabase

📊 Complete Feature List

Data Generation

• Schema-based generation (no training data required)
• Dataset-based ML generation (CTGAN, TVAE, GaussianCopula)
• LLM-powered seed data generation
• Differential privacy with configurable ε/δ
• DP parameter validation & recommendations
• Model download & export

Data Management

• CSV upload with auto-profiling
• Schema detection & type inference
• PII/PHI column detection
• Distribution analysis & statistics
• Correlation matrices
• Missing value analysis

Quality & Privacy

• Statistical similarity scoring
• ML utility evaluation (classification/regression)
• Privacy risk assessment
• Membership inference testing
• k-anonymity checks
• Privacy budget tracking

AI Assistant

• Natural language queries
• Context-aware recommendations
• Code generation for API usage
• Error debugging
• Compliance guidance

Enterprise

• HIPAA/GDPR/SOC-2 compliance reports
• Immutable audit logs
• Usage & billing dashboards
• Role-based access control
• OAuth (Google, GitHub)
• Two-factor authentication

🗺️ Roadmap

• FHIR/HL7 medical data formats
• Time-series synthetic data
• Enterprise SSO (SAML 2.0)
• Python & JavaScript SDKs
• Self-hosted Docker templates
• Real-time streaming generation

See CHANGELOG.md for version history.