A financial crime investigation platform for detecting suspicious transactions, analyzing financial networks, and generating regulatory reports.
FinSentry converts raw transaction datasets into structured investigation intelligence using fraud detection models, explainable analytics, graph-based investigation tools, and automated SAR report generation.
Financial institutions process millions of transactions daily, making fraud investigation and AML monitoring difficult.
FinSentry provides a complete investigation pipeline that allows analysts to:
• detect suspicious transactions • understand fraud model explanations • explore transaction networks visually • identify money laundering patterns • build investigation cases • generate Suspicious Activity Reports (SAR)
The platform provides an interactive investigation dashboard for financial crime analysts.
The dashboard allows investigators to upload transaction datasets and run the full investigation pipeline.
The investigation page visualizes the transaction network, allowing analysts to explore entity relationships, suspicious flows, and risk scores.
Generated SAR reports summarize suspicious activity, evidence, and regulatory insights.
Transaction Dataset
│
▼
Fraud Detection
(Isolation Forest + Random Forest)
│
▼
Explainability
(SHAP Feature Attribution)
│
▼
Graph Intelligence
(NetworkX Transaction Graph)
│
▼
AML Pattern Detection
│
▼
Case Builder
│
▼
Investigation Narrative
│
▼
SAR Generator
│
▼
Compliance Validator
│
▼
Investigation Dashboard
Transactions are analyzed using machine learning models:
• Isolation Forest • Random Forest
Each transaction receives a fraud probability score and risk classification.
FinSentry integrates SHAP to explain fraud predictions.
Investigators can view:
• feature importance • anomaly drivers • entity-level risk explanations
This enables transparent and auditable decision-making.
Financial crime often occurs across networks of transactions.
FinSentry builds a transaction graph where:
• nodes represent entities • edges represent financial transfers
Graph analytics include:
• degree centrality • PageRank • community detection • cycle detection • suspicious transfer chains
The system automatically detects common AML typologies:
• structuring • layering • round-tripping • rapid transfers • shell company clusters
Each detection includes investigator-friendly explanations.
Suspicious transactions are grouped into investigation cases containing:
• involved entities • suspicious transactions • AML patterns detected • evidence and indicators • narrative explanations
FinSentry generates Suspicious Activity Reports automatically.
Reports include:
• suspicious activity description • transaction evidence • risk assessment • jurisdiction analysis • entities involved
Generated SAR reports are stored and retrievable.
Available endpoints:
| Endpoint | Description |
|---|---|
GET /sar/list |
List stored SAR reports |
GET /sar/{report_id} |
Retrieve SAR report |
This allows investigators to review historical reports.
Upload transaction datasets.
Displays suspicious transactions with fraud scores.
Interactive network graph showing suspicious transaction flows.
Shows entity risk metrics and connected entities.
Displays investigation cases and AML patterns.
Generate and browse SAR reports.
Python FastAPI Pydantic scikit-learn SHAP NetworkX SQLAlchemy
React TypeScript Vite TailwindCSS react-force-graph
pytest
326 automated tests
| Endpoint | Description |
|---|---|
POST /pipeline/run |
Run investigation pipeline |
GET /case/{case_id} |
Retrieve investigation case |
GET /timeline/{case_id} |
Investigation timeline |
GET /graph/{entity_id} |
Graph metrics for entity |
GET /entity/{entity_id} |
Entity investigation profile |
POST /sar/generate |
Generate SAR report |
GET /sar/list |
List stored SAR reports |
GET /sar/{report_id} |
Retrieve SAR report |
FinSentry/
│
├── agents/
├── aml_patterns/
├── case_builder/
├── compliance_validator/
├── explainability/
├── fraud_detection/
├── graph_engine/
├── graph_rag/
├── ingestion/
├── investigation_narrative/
├── sar_generator/
├── api/
│
├── frontend/
│
├── data/
│ └── sample_data_transactions.csv
│
├── screenshots/
│ ├── dashboard.png
│ ├── investigation.png
│ └── report.png
│
├── tests/
└── requirements.txt
Clone the repository
git clone https://github.com/parth-shinge/FinSentry.git
cd FinSentryCreate virtual environment
python -m venv venvActivate environment
Mac/Linux
source venv/bin/activateWindows
venv\Scripts\activateInstall dependencies
pip install -r requirements.txtStart the API server
uvicorn api.main:app --reloadBackend runs at
http://localhost:8000
API docs
http://localhost:8000/docs
cd frontend
npm install
npm run dev
Frontend runs at
http://localhost:5173
1. Upload sample transaction dataset
2. Run investigation pipeline
3. Explore transaction graph
4. Investigate entities and cases
5. Generate SAR report
6. Review stored SAR reports
✔ Explainable fraud detection ✔ Graph-based AML investigation ✔ Automated case generation ✔ SAR report generation and storage ✔ Interactive investigation dashboard
FinSentry demonstrates how fraud detection models, explainable analytics, and graph-based investigation tools can be integrated into a unified system for financial crime analysis and AML compliance.
The platform enables investigators to identify suspicious networks, build investigation cases, and generate regulatory reports efficiently.