How vijayaramanujam/hack-with-adithya works
This repository detects suspected money-mule accounts, combines ML and rule signals into account risk scores, and identifies connected fraud rings. The principal workflow is CSV loading, validation, preprocessing, graph and behavioral feature generation, classification and rule detection, risk fusion, then ring analysis. The batch CLI orchestrates training and inference; a FastAPI layer serves predictions and saved analysis outputs. The graph preserves README-described stages and API capabilities. Some relationships rely on the shown pipeline references or README rather than complete source excerpts; unshown implementation details are not inferred.
Main entry points: train.py Training workflow · predict.py Scoring workflow · preprocessing.py Preprocessing · inference.py Model inference
Components 20
Data preparation
- Dataset loader
src/data_loader.py - Data validation
src/data_validation.py - Preprocessing
src/preprocessing.py
Detection analysis
- Transaction graph
src/graph_builder.py - Account features
src/feature_engineering.py - Rule detectors
src/rule_detectors.py - Risk fusion
src/risk_scoring.py - Fraud ring detection
src/ring_detection.py
Model training
- Mule classifier
src/train_model.py - Model evaluation
src/model_evaluation.py - Training workflow
train.py
Inference and API
- Model inference
src/inference.py - Scoring workflow
predict.py - FastAPI service
api/main.py - API schemas
api/schemas.py
Configuration and artifacts
- Pipeline configuration
src/config.py - Model artifacts
models
Other components
- Analyst / operator
- Financial CSV data
- API client
Connections 41
- Financial CSV data → Dataset loader: loads
README.md - Dataset loader → Data validation: validates
train.py - Data validation → Preprocessing: cleans
train.py - Preprocessing → Transaction graph: builds topology
README.md - Preprocessing → Account features: derives features
README.md - Transaction graph → Account features: adds graph metrics
README.md - Preprocessing → Rule detectors: checks transactions
predict.py - Account features → Mule classifier: trains on features
train.py - Rule detectors → Mule classifier: adds rule signals
train.py - Mule classifier → Model evaluation: evaluates
train.py - Account features → Model inference: scores features
predict.py - Rule detectors → Risk fusion: supplies rule scores
predict.py - Model inference → Risk fusion: supplies probabilities
predict.py - Preprocessing → Risk fusion: supplies transactions
predict.py - Risk fusion → Fraud ring detection: ranks suspicious accounts
predict.py - Training workflow → Dataset loader: loads data
train.py - Training workflow → Data validation: validates data
train.py - Training workflow → Preprocessing: preprocesses data
train.py - Training workflow → Transaction graph: builds graph
train.py - Training workflow → Account features: builds features
train.py - Training workflow → Rule detectors: runs detectors
train.py - Training workflow → Mule classifier: trains model
train.py - Training workflow → Model evaluation: evaluates model
train.py - Training workflow → Risk fusion: computes risk
train.py - Training workflow → Fraud ring detection: detects rings
train.py - Training workflow → Model inference: uses inference
train.py - Scoring workflow → Dataset loader: loads data
predict.py - Scoring workflow → Preprocessing: preprocesses data
predict.py - Scoring workflow → Transaction graph: builds graph
predict.py - Scoring workflow → Account features: builds features
predict.py - Scoring workflow → Rule detectors: runs detectors
predict.py - Scoring workflow → Model inference: predicts probabilities
predict.py - Scoring workflow → Risk fusion: computes risk
predict.py - Scoring workflow → Fraud ring detection: detects rings
predict.py - API client → FastAPI service: requests endpoints
api/main.py - FastAPI service → API schemas: uses response models
api/main.py - FastAPI service → Model inference: predicts
api/main.py - Pipeline configuration → Dataset loader: sets data paths
src/data_loader.py - Pipeline configuration → Preprocessing: sets output paths
src/preprocessing.py - Pipeline configuration → Model inference: locates artifacts
src/inference.py - Model inference → Model artifacts: loads model metadata
src/inference.py