Financial institutions spend millions on proprietary compliance solutions, locked into expensive vendor ecosystems. When regulators send ad hoc inquiries via email/PDF, legal teams face manual, error-prone processes taking weeks to interpret requests & compile responses from multiple data sources.
We present open source, agentic AI platform for regulatory compliance using NLP & bitemporal data.
Core architecture: XTDB provides immutable bitemporal storage for audit trails, answering "What did you know and when?" Open LLMs via Ollama enable natural language understanding, with FSI terminology dictionaries (NAICS, GICS). LangChain agents translate compliance questions to precise queries.
"Reason and Act" architecture Reason Agent interprets inquiries & identifies data sources; Query Agent generates validated queries & synthesizes regulatory-compatible responses. Stack: Kubernetes, OpenFaaS, Kafka, S3, HashiCorp Vault, OpenTelemetry, Redis, Spark, Keycloak, NiFi,Kubeflow.
AWS ref architecture with optional managed services (RDS, MSK, EKS) for faster deployment.
Attendees get blueprint for building AI systems in regulated environments using open source components, full audibility.