Skip to content

Regulatory Compliance

Compliance work often comes down to one question: is this entity connected, directly or indirectly, to something it should not be, like a sanctioned entity? Odin is well-suited to it because the prohibited relationships are frequently indirect, hidden a few hops from anything an analyst would look at directly.

The graph holds entities like Fund, Manager, Entity, Person, and Jurisdiction, connected by relations such as managed_by, owns, affiliated_with, and controlled_by.

The direct case: check one edge

When you have a specific prohibited relationship in mind, score it directly, with no traversal needed:

from arango import ArangoClient
from odin import OdinEngine

db = ArangoClient(hosts="http://localhost:8529").db(
    "compliance", username="root", password=""
)
engine = OdinEngine(db, community_id="compliance", community_mode="mapping")

score = engine.score_edge(
    "entity/investment_fund",
    "managed_by",
    "entity/sanctioned_entity",
)
if score > 0.5:
    compliance_agent.flag_for_review(
        "Fund plausibly managed by a sanctioned entity"
    )

score_edge() gives you the NPLL plausibility of a single relationship, a fast gate for rule checks. See Scoring Edges.

Find indirect exposure with retrieval

Direct edges are the easy case. To catch indirect ties, retrieve paths from the fund and look for any that reach a sanctioned entity:

result = engine.retrieve(
    seeds=["entity/investment_fund"],
    hop_limit=4,          # indirect control can be several hops away
    max_paths=100,
)

for p in result["paths"]:
    node_ids = [e["u"] for e in p["edges"]] + [e["v"] for e in p["edges"]]
    if any("sanctioned" in str(n) for n in node_ids):
        compliance_agent.flag_for_review(p)

This finds chains like Fund → managed_by → Manager → affiliated_with → Sanctioned Entity that a single-edge check would miss.

Why NPLL helps compliance

Concern How Odin helps
False positives from coincidental edges NPLL down-weights implausible connections
Hidden, indirect relationships Multi-hop beam search surfaces them
Explainability for auditors Every flagged path is a real, inspectable chain in the graph

Because Odin only returns paths that exist in your data, every flag is auditable: there is no hallucinated relationship to defend.

Escalation gate

score = result["triage"]["score"]
if score >= 75:
    compliance_agent.escalate(result)     # strong, well-sourced exposure
elif score >= 40:
    human_review.enqueue(result)          # uncertain, needs an analyst

See AI Agent Integration for the human-in-the-loop pattern.

Next