All projects
AI & ML
Financial Services
Vector Bank Fraud Graph
Real-time graph-based fraud detection at a tier-one bank.
Client
Vector Bank
Delivered
2025 · 6 months
Team
8 specialists
Category
AI & ML
Overview
What we built
Delivered a real-time fraud detection platform combining a Neo4j graph, streaming features and gradient-boosted models scoring 3,500 transactions per second.
Business impact
Outcomes that moved the P&L
0%
Fraud losses
0%
False positives
0
Peak TPS scored
₹0 Cr
Savings / year
The challenge
- Rule-based fraud engine missed emerging synthetic-identity patterns.
- Investigators worked across five disconnected tools with no shared context.
- Regulator required explainable decisions for every declined transaction.
How we approached it
- Built a customer-merchant-device graph in Neo4j updated from Kafka in near real time.
- Combined graph features with XGBoost models and SHAP-based explanations.
- Shipped a case-management workspace giving investigators one view of the network.
What changed
- Fraud losses reduced by 32% in the first six months post-launch.
- False positives cut in half, restoring 1.1M legitimate transactions per year.
- Every model decision now ships with a regulator-ready explanation.
Project gallery
Inside the delivery
Streaming feature store
Sub-100ms feature retrieval.
Investigator workspace
One canvas for the entire fraud network.
Model governance
Every decision explainable end-to-end.
Technology
Built with
The tools and platforms we chose for this engagement.
Neo4j
Apache Kafka
Python
XGBoost
MLflow
AWS EKS
Feast
"Dataamps brought both the ML depth and the banking discipline our regulators expected. It's the strongest partnership we've had."
Marcus Weller — Chief Risk Officer, Vector Bank
Talk to our team
Have a similar challenge?
Tell us about your goals. We'll put together a scoped conversation with the right specialists.
Ready to amplify your data?
Book a 30-minute strategy call. We'll map your data landscape and identify the three highest-leverage moves.