AI Agents for Money Laundering Detection
This console tracks the project across its three build phases — from a first chat-driven agent through richer logging and oversight, up to more advanced reasoning. Each phase connects to the same underlying research pipeline, with the live agent responses served from AWS Bedrock.
Foundation Agent
A single conversational agent with access to case records and dataset statistics. This was a test and setup for Phase 2 and is now redundant, so it is no longer needed.
Agent Logging & Oversight
Builds on Phase 1 with a live log panel exposing the agent's reasoning and tool calls step by step, so a human reviewer can audit how it reached a conclusion.
Supervisor Evaluation
The same investigation pipeline as Phase 2, run as a training case in which the agents make deliberate mistakes. Question their answers, then ask the trainer to score how you worked with them.