PTAB

IPR2025-01359

Zestyai Inc v. Aon Re Inc

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: System for Automated Property Feature Assessment
  • Brief Description: The ’058 patent discloses systems and methods for automatically assessing property features by applying machine learning analysis to aerial imagery. The technology is used to extract characteristics like building materials and classify conditions such as damage for risk evaluation purposes.

3. Grounds for Unpatentability

Ground 1: Claims 1-20 are obvious over Gross.

  • Prior Art Relied Upon: Gross (Application # 2015/0186953).
  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner argued that Gross, a single prior art reference not considered during prosecution, discloses every limitation of the challenged claims. Gross teaches a “Classifier Engine” that is trained using reference images to assess real property features and their conditions, directly corresponding to the ’058 patent’s use of machine learning. The engine is trained with images of building elements to identify their “types” (e.g., shingle roof) and “conditions” (e.g., damaged), which Petitioner contended meets the limitations of developing “property characteristic profiles” and “property condition profiles.” Gross explicitly discloses using neural networks and incorporates by reference the Jafri publication, which details using convolutional neural networks (CNNs) for image analysis. Petitioner asserted this disclosure renders claims reciting CNNs (claim 9) and deep learning algorithms (claim 11) obvious.

    • For independent claims 1 and 13, Petitioner mapped Gross’s system step-by-step. Gross’s engine accesses property images (which can be aerial images), identifies features (e.g., roof), determines a plurality of their characteristics (e.g., shape and material), classifies their condition, and presents the results to a user in a graphical interface. For dependent claims, Petitioner argued Gross further teaches specific features like approximating the size of a roof area (claim 12) and performing “long term evaluations” by comparing current assessments to stored historical data (claim 10).

    • Motivation to Combine: The petition asserted that a person of ordinary skill in the art (POSA) would find it obvious to apply the specific image processing and machine learning techniques Gross describes for training its Classifier Engine to the subsequent assessment process performed by that same trained engine. Petitioner argued this is the entire point of Gross’s system—to use the trained classifier to assess new properties. For claim elements requiring receipt of a request for a specific property location, Petitioner argued a POSA would combine Gross’s disclosure of a user interface with this functionality. A POSA would modify Gross’s interface from a broad geographic search to a specific property request to improve efficiency and provide on-demand analysis, which was a well-known and predictable design choice.

    • Expectation of Success: A POSA would have had a clear expectation of success in implementing Gross’s system as claimed. The proposed application involves using conventional and well-understood image processing and machine learning techniques for their intended purpose, which would yield predictable results in property feature classification.

    • Key Aspects: Petitioner emphasized that the ’058 patent issued without a single rejection based on prior art under 35 U.S.C. §102 or §103. The core argument is that the patent claims nothing more than the routine application of well-established machine learning concepts to the long-standing practice of property assessment, a combination fully disclosed by Gross.

4. Relief Requested

  • Petitioner requests the institution of an inter partes review and the cancellation of claims 1-20 of the ’058 patent as unpatentable.