Skip to content
Hari.Back to projects

Case studyAI / ML

ReconAI

Trace invoice mismatches and vendor risk through graphs.

Year
2026
Focus
Graph-based GST reconciliation

Overview

A GST reconciliation engine using knowledge graphs for invoice matching, circular-trade detection, vendor risk, and audit trails.

The challenge

GST reconciliation involves invoice mismatches, input-tax-credit chains, circular transactions, and risk across connected vendors.

How it works

  1. Model tax records in Neo4j and reconcile invoices, input-tax-credit chains, circular trades, and propagated risk.

  2. Extract graph features for XGBoost vendor-risk predictions with SHAP explanations.

  3. Expose FastAPI endpoints to a React dashboard and generate audit reports with traversal paths and regulatory references.

Technology stack

  • Python
  • FastAPI
  • Neo4j
  • NetworkX
  • XGBoost
  • SHAP
  • React

Highlights

GST reconciliation
Graph
vendor scores with SHAP
Risk
explainable report trails
Audit

Let’s talk about this project.

Discuss this project