ML Code Release Authorization With Adaptive Testing Rules

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional software development and release processes require manual coordination with security teams and rely on inefficient, one-size-fits-all testing, leading to unnecessary computing resource consumption and inefficiencies.

Innovation Solution

An agile developer portal utilizing a machine learning-based system that autonomously develops, tests, and determines code release authorization, customizing testing and evaluation based on release type and impact, with a rules engine and machine learning model to make intelligent release decisions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional manual coordination with security teams and standard testing schemes are used, then code release can be controlled and security requirements met, but the process consumes unnecessary computing processing power and is inefficient

Engineering Contradiction:

Improvecode release efficiencyVSAvoidcomputing processing power consumption

Core Design Contradiction:

ProductivityVSUse of energy by moving object

Solution Approach 1:

The system dynamically changes testing parameters by analyzing code characteristics, release type, and impact scope to determine which tests are necessary. This avoids running all available tests uniformly, thereby reducing unnecessary computing resource consumption while maintaining adequate quality control.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The machine learning model autonomously evaluates code and determines release authorization without requiring manual coordination with security teams. The system self-manages the release decision process by analyzing code properties and comparing them against learned patterns from historical data, eliminating inefficient human-in-the-loop coordination.

Inventive Principle:

Principle #25Self-service

2Reliability

If all available tests are run for every code release, then comprehensive code quality assurance is achieved, but unnecessary computing processing power is consumed when not all tests are required

Engineering Contradiction:

Improvecode quality assuranceVSAvoidcomputing processing power

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system applies different testing strategies based on local characteristics of the code being released. By analyzing specific code properties, release type, and impact scope, the system determines which subset of tests is appropriate for each release scenario, avoiding the waste of running all tests uniformly across different code changes.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

Instead of always performing the complete set of available tests (excessive action), the machine learning model determines the partial set of tests that are actually necessary for each release. This partial action approach maintains adequate quality assurance while avoiding unnecessary computing resource consumption.

Inventive Principle:

Principle #16Partial or excessive action

3Reliability

If manual coordination processes are used for code release, then regulatory requirements can be met, but the process is inefficient and time-consuming

Engineering Contradiction:

Improveregulatory complianceVSAvoidrelease coordination time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system replaces manual mechanical coordination processes with an automated machine learning-based decision system. The ML model analyzes code properties, release types, and impact scopes to automatically determine release authorization, substituting the slow manual coordination process while maintaining regulatory compliance through systematic evaluation.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system implements automated feedback loops where the machine learning model continuously evaluates code against learned patterns from historical release data and regulatory requirements. This automated feedback mechanism replaces manual coordination while ensuring compliance, as the model learns from past decisions and applies consistent evaluation criteria.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12530188B2Machine learning-based system for code development and release

Publication Date: 2026.01.20 BANK OF AMERICA CORP

AI summary

Arrangements for machine learning-based code development and release are provided. In some examples, a request to release code may be received from a developer portal and may include the code and a type of release. Based on the type of release, one or more requirements for release may be identified by a rules engine. Code testing data responsive to the one or more requirements may be retrieved from the developer portal and a machine learning model may be executed using the code, type of release and code testing data as inputs. The model may output a release authorization decision. If the release authorization decision includes approval of the release, the code may be deployed to a production environment. If the release authorization decision includes a denial of the request to release the code, the code may be returned to the developer portal for further modification or testing.