Fighting Corruption with Technology
summary
Claims that current systems catch only 1–3% detection today and therefore fail, and that a scalable pipeline named ComplyDetect pipeline builds a usable alternative. System downloads the web with Hadoop crawlers, streams pages into Kafka, runs NER and three ML detectors (politician, relationship, crime) trained via active sampling at decision boundaries, stores results in MongoDB and Elasticsearch, and processes 20 million documents a day—about 800× the throughput of manual analysts—intended to surface laundering links for onboarding and transaction screening.
AI-written summary · about
chapters (16)
- 0:00Introduction by host
- 1:02Can technology fight corruption
- 2:26Corruption's human cost
- 4:32Scale of money laundering
- 6:47Politicians and financial risk
- 8:11Networks of corruption
- 9:15Problem for banks today
- 11:02Public data challenge
- 12:24Building a tech solution
- 13:25Hadoop for big data
- 14:10Kafka for streaming data
- 14:51Named entity recognition
- 15:36Machine learning detectors
- 17:00MongoDB and Elasticsearch
- 17:44Training machine learning models
- 19:08Scaling with technology
related talks
24mHunting Criminals with Hybrid Analytics
25mReal-Time Fraud Detection at Scale
51mReal-time Fraud Detection: Challenges and Solutions
10mReal-Time Fraud Prevention at Scale
30mProcessing Threat Reports at Scale Using AI and ML: Expectations and Reality
16mLeveraging ML to Create a Reliable Anti-Fraud System
08mAI-Powered Fraud Detection in Financial System
42mBuilding a Robust Fraud Detection System