Anti-Money Laundering AI Solution
An AI-based solution that detects suspicious behaviors in large-scale financial transactions.
Business issue
Why I was brought into the project
Compliance teams need to detect suspicious behaviors in large volumes of financial transactions while staying aligned with AML, KYC and regulatory expectations.
Context
The environment around the project
Compliance teams need tooling to detect and investigate suspicious behaviors across financial transactions.
Functional environment
The functional context involved anti-money laundering monitoring, suspicious behavior detection and compliance investigation support.
Technical environment
The technical environment combined Dataiku, Python, machine learning, transaction analytics and centralized dashboards.
Challenges
The solution had to surface meaningful anomalies, support investigation workflows and remain understandable for compliance users.
Solution
My contribution and its impact
My contribution to the project
I designed and implemented an AI-based solution to detect suspicious transaction behaviors and delivered a centralized dashboard for investigation and monitoring.
- AI-based suspicious behavior detection workflow
- Centralized AML investigation dashboard
- KYC and FATF-aligned solution structure
- Operational monitoring views for compliance teams
Impact
The solution improved visibility on potentially suspicious activity and gave compliance teams a more operational way to investigate anomalies.
- Better visibility on suspicious transaction patterns
- More structured investigation support
- Improved operational effectiveness for compliance users
Impact metrics
Approach
How I structured the work
- Frame AML detection needs and relevant behavioral signals.
- Build detection logic and analytics workflows in Dataiku.
- Design dashboard views that help compliance teams investigate suspicious activity.
Takeaways
What I learned from this project
- In compliance use cases, explainability and usability are as important as detection performance.
- Dashboards must support investigation, not only display model outputs.
- Regulatory alignment shapes both the model design and the user experience.