AI Agent Routing via Semantic Fingerprinting and Bloom Filters

ChangeBridge: Patent Grants - Networking (H04L)

Published March 31st, 2026

Detected March 31st, 2026

Summary

The USPTO granted Patent US12591793B2 to Citibank, N.A. covering a hierarchical semantic fingerprinting system for AI agent routing that uses locality-sensitive hash functions and bloom filter cascades. The patent discloses methods for matching queries to appropriate AI agents by comparing semantic fingerprints.

View original document View source feed page

What changed

USPTO granted Patent US12591793B2 to Citibank N.A. on March 31, 2026. The patent discloses systems and methods for orchestrating task execution among autonomous AI agents using hierarchical semantic fingerprinting, locality-sensitive hashing (LSH) functions, and a bloom filter cascade to match queries with agent capabilities. Inventors include Ganesh Prasad Bhat, James Myers, Sourabh Deb, Jason Engelbrecht, Zheyu Wang, and Haolin Jin. The system generates semantic-aware fingerprints for queries and agents, with subsequent hierarchical layers encoding more complex semantic information.

Companies developing AI agent routing systems should review this patent for potential licensing needs or freedom-to-operate analysis. While patent grants do not impose compliance deadlines, organizations implementing AI agent systems should assess whether their technologies fall within the scope of this intellectual property. Legal teams should evaluate the claims to determine if development activities may implicate this patent.

Source document (simplified)

← USPTO Patent Grants

Hierarchical cascade architecture of semantic fingerprinting operations for agent routing

Grant US12591793B2 Kind: B2 Mar 31, 2026

Assignee

Citibank, N.A.

Inventors

Ganesh Prasad Bhat, James Myers, Sourabh Deb, Jason Engelbrecht, Zheyu Wang, Haolin Jin

Abstract

The systems and methods disclosed herein orchestrate task execution among autonomous (or semi-autonomous) AI agentic models (“agents”) responsive to a received query by using a hierarchical semantic fingerprinting framework to generate semantic-aware fingerprints for the query and agents. Queries and descriptions of agent capabilities are processed by a series of hierarchical levels using locality-sensitive hash (LSH) functions, where subsequent layers encode more complex semantic information and generate longer hash values. The hash values are aggregated into a semantic fingerprint. A bloom filter cascade uses a series of increasingly accurate hierarchical bloom filters to reject agent fingerprints that differ from the query fingerprint. The remaining agent fingerprints are compared bitwise to the query fingerprint, and those closest to the query fingerprint are selected to generate a routing path for the query. Responses from the selected agents are aggregated into an output that is responsive to the input.