LLM Enabled Multi Document Correlation for Financial Compliance
summary
Claims a working pipeline that ingests ~3 million records across four jurisdictions and five years and combines a graph-based entity correlation engine, an adaptive probabilistic risk model, jurisdictional normalization, and an LLM reasoning layer to surface cross‑document compliance risks. Built system detects multi‑system patterns missed by rules, cuts false positives, and reduces auditor effort in evaluations; challenges include LLM hallucination, data quality, scalability, governance, and the need for human-verified evidence.
AI-written summary · about
chapters (16)
- 0:00Welcome and Background
- 0:30Why Compliance Risks Hide
- 2:13Talk Roadmap
- 4:24Why Rules Fall Short
- 6:04Three-Part AI Framework
- 7:52Pipeline and Feedback Loop
- 11:10Business Value of AI
- 12:57Real-World Scenario Walkthrough
- 14:49Graph Correlation Engine
- 16:38Probabilistic Risk Scoring
- 18:21Jurisdiction Normalization
- 19:50LLMs for Explainability
- 21:41Evaluation Setup and Results
- 24:54From Reactive to Predictive
- 26:35Deployment and AI Challenges
- 29:22Key Takeaways and Closing
related talks
24mLLMs in Regulated Industries: Challenges and Governance Solutions
44mGive Your LLMs a Left Brain
13mLLM-Audited EHR Pipelines: 87% Accuracy with 70% Fewer Resources
47mGive Your LLMs a Left Brain
33mDeep research at scale
24mConversational Search with LLMs
13mHarmonising LLM Agents for Real-World Market Intelligence:[...]
40mLLMs & Data Warehouses: Reliable AI Agents for Business (Keynote)