AML-Nexus — Catch the Criminals. Protect Everyone Else.
Inspiration
What if AI could detect financial crime without unnecessarily harming innocent people?
Behind every financial account is a real person.
Imagine a small business owner whose account is restricted after legitimate customer payments are mistaken for suspicious activity. Suddenly, paying employees, covering rent, and keeping the business running could become difficult.
Or imagine a charity receiving donations from hundreds of people, only to have legitimate fundraising activity resemble suspicious fund aggregation.
These scenarios inspired the central question behind AML-Nexus:
Can we build financial crime detection that is both intelligent and fair?
AML-Nexus was created to demonstrate that identifying suspicious financial networks and protecting innocent customers should be complementary goals.
What It Does
AML-Nexus is an explainable AI-powered anti-money-laundering investigation copilot.
It helps investigators identify suspicious transaction networks, understand supporting and contradicting evidence, and make proportionate decisions with human oversight.
Intelligent Money-Flow Detection
AML-Nexus analyzes transaction networks to identify patterns such as fund aggregation, circular transfers, and connected suspicious accounts.
Interactive network graphs and transaction timelines make complex money movements easier to investigate.
Evidence That Works Both Ways
Not every unusual transaction is criminal.
AML-Nexus examines both evidence supporting suspicion and evidence that may indicate legitimate activity.
This helps investigators understand the reasons behind recommendations rather than relying on opaque alerts.
Human-Controlled Decisions
The platform supports four graduated investigation recommendations:
- Clear: No further action indicated.
- Verify: Request additional information.
- Monitor: Continue evidence-based observation.
- Escalate: Refer the case for urgent human review.
AI recommends. Humans decide.
Silent Evidence Watch
AML-Nexus includes a structured monitoring workflow that allows investigators to gather additional evidence without automatically restricting customer accounts.
This is especially valuable when suspicious-looking activity may have a legitimate explanation.
PayPal Sandbox Integration
AML-Nexus integrates with PayPal Sandbox using OAuth 2.0 and the Transaction Search API.
Completed sandbox transactions are retrieved through a read-only connection and normalized for analysis using the same detection pipeline as the demonstration scenarios.
Google Gemini Investigation Summaries
Google Gemini helps generate investigation summaries grounded in structured case evidence.
Numerical validation checks and fallback handling help keep the investigation workflow reliable when AI-generated summaries are unavailable.
Explainable Statistical Intelligence
A Naive Bayes evidence ledger with Laplace smoothing provides interpretable statistical assessments.
The application also incorporates transaction-pattern analysis and statistical indicators to support investigation.
Auditability and PDF Reports
AML-Nexus includes reviewer-accountability controls and tamper-evident hash-chain audit records.
Investigators can also
Built With
- anti-money-laundering
- explainable-ai
- fintech
- github-actions
- google-gemini
- machine-learning
- naive-bayes
- networkx
- numpy
- oauth2
- pandas
- paypal-rest-api
- paypal-sandbox
- python
- regtech
- streamlit
