# US 2026/0227960 A1 · Self-evolving infrastructure using llm

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0227960 A1
- Document ID: US20260227960A1
- Status: Published application
- Kind: A1
- Application number: 19/046,315
- Filed: 2025-02-05
- Published: 2026-08-06
- Earliest claimed: 2025-02-05
- Applicants: Oracle International Corporation
- Inventors: Paul Gregory GREENSTEIN; Zbigniew HOLKA
- Web page: https://patentgrep.com/patent/US20260227960A1

## CPC

- G06F 8/10: Requirements analysis; Specification techniques
- G06F 8/60: Software deployment
- G06F 8/61: Installation
- G06F 8/65: Updates (security arrangements therefor G06F21/57)
- G06N 3/0475: Generative networks
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Systems, methods, and other embodiments associated with self-evolution of compute infrastructure using LLMs are described. In one embodiment, a method includes monitoring a parameter relating to a compute infrastructure that is deployed to a target computing system for a condition. In response to detection of the condition, the method revises infrastructure requirements for the compute infrastructure to express an adaptation to the condition using a first LLM. The infrastructure requirements are in human language. The method modifies an infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using one or more additional LLMs. The method converts the modified infrastructure topology into an executable deployment specification for evolved compute infrastructure that conforms to the revised infrastructure requirements. And, the method executes the deployment specification to install the evolved compute infrastructure on the target computing system to cause the parameter to cease to detect the condition.

## Claims

### Claim 1

One or more non-transitory computer-readable media that include stored thereon computer-executable instructions that when executed by at least a processor of a computing system cause the computing system to:
monitor a parameter relating to a compute infrastructure that is deployed to a target computing system for a condition;
in response to detection of the condition, revise infrastructure requirements for the compute infrastructure to express an adaptation to the condition using a first LLM, wherein the infrastructure requirements are in human language;
modify an infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using one or more additional LLMs;
convert the modified infrastructure topology into an executable deployment specification for evolved compute infrastructure that conforms to the revised infrastructure requirements; and
execute the deployment specification to install the evolved compute infrastructure on the target computing system to cause the parameter to cease to detect the condition.

### Claim 2 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, further comprising instructions that when executed by at least the processor cause the computing system to:
return to monitoring the parameter until an instruction to terminate the monitoring is received; and
in response to a further detection of the condition, repeat the steps of revising the infrastructure requirements, modifying the infrastructure topology, converting the modified infrastructure topology into the deployment specification, and executing the deployment specification to install the evolved compute infrastructure on the target computing system.

### Claim 3 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, further comprising instructions that when executed by at least the processor cause the computing system to:
detect the condition because the compute infrastructure is underutilized based on the parameter, wherein the parameter indicates a level of usage of the compute infrastructure; and
use the first LLM to revise the infrastructure requirements to reduce compute resources of the compute infrastructure.

### Claim 4 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, further comprising instructions that when executed by at least the processor cause the computing system to:
detect the condition because the compute infrastructure is overutilized based on the parameter, wherein the parameter indicates a level of usage of the compute infrastructure; and
use the first LLM to revise the infrastructure requirements to increase compute resources of the compute infrastructure.

### Claim 5 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, further comprising instructions that when executed by at least the processor cause the computing system to:
detect the condition because an update to a component of the compute infrastructure has become available; and
use the first LLM to revise the infrastructure requirements to specify the component of the compute infrastructure to be configured according to the update.

### Claim 6 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, wherein the computer-executable instructions to modify the infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using the one or more additional LLMs, when executed by at least the processor, further cause the computing system to:
translate the revised infrastructure requirements into a current logical infrastructure topology using one of the additional LLMs that is trained to generate graphs of logical infrastructure from the infrastructure requirements and a prior logical infrastructure topology; and
translate the revised infrastructure requirements and the current logical infrastructure topology into a physical infrastructure topology using one of the additional LLMs that is trained to generate graphs of physical infrastructure from infrastructure requirements, the current logical infrastructure topology, and a prior physical infrastructure topology, wherein the physical infrastructure topology is used as the modified infrastructure topology.

### Claim 7 (depends on 1)

The one or more non-transitory computer-readable media of claim 1, wherein the executable deployment specification is written in YAML code or HCL code.

### Claim 8

A computer-implemented method, the method comprising:
monitoring a parameter relating to a compute infrastructure that is deployed to a target computing system for a condition;
in response to detection of the condition, revising infrastructure requirements for the compute infrastructure to express an adaptation to the condition using a first LLM, wherein the infrastructure requirements are in human language;
modifying an infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using one or more additional LLMs;
converting the modified infrastructure topology into an executable deployment specification for evolved compute infrastructure that conforms to the revised infrastructure requirements; and
executing the deployment specification to install the evolved compute infrastructure on the target computing system to cause the parameter to cease to detect the condition.

### Claim 9 (depends on 8)

The computer-implemented method of claim 8, further comprising performing the steps of claim 8 in a loop that iterates each time the condition is detected until an instruction to terminate the monitoring is received.

### Claim 10 (depends on 8)

The computer-implemented method of claim 8, further comprising:
detecting the condition because the compute infrastructure is overutilized or underutilized based on the parameter, wherein the parameter is a level of usage of the compute infrastructure; and
using the first LLM to revise the infrastructure requirements to adjust compute resources of the compute infrastructure.

### Claim 11 (depends on 8)

The computer-implemented method of claim 8, further comprising:
detecting the condition because an update to a component of the compute infrastructure has become available; and
using the first LLM to revise the infrastructure requirements to specify the component of the compute infrastructure to be configured according to the update.

### Claim 12 (depends on 8)

The computer-implemented method of claim 8, wherein modifying the infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using the one or more additional LLMs further comprises:
translating the revised infrastructure requirements into a current logical infrastructure topology using one of the additional LLMs that is trained to generate graphs of logical infrastructure from the infrastructure requirements and a prior logical infrastructure topology; and
translating the revised infrastructure requirements and the current logical infrastructure topology into a physical infrastructure topology using one of the additional LLMs that is trained to generate graphs of physical infrastructure from infrastructure requirements, the current logical infrastructure topology, and a prior physical infrastructure topology, wherein the modified infrastructure topology is the physical infrastructure topology.

### Claim 13 (depends on 8)

The computer-implemented method of claim 8, further comprising generating the executable deployment specification in YAML code or HCL code.

### Claim 14

A computing system, comprising:
at least one processor connected to at least one memory;
one or more non-transitory computer-readable media that include stored thereon computer-executable instructions that when executed by at least the processor accessing the memory cause the computing system to:
monitor a parameter relating to a compute infrastructure that is deployed to a target computing system for a condition;
in response to detection of the condition, revise infrastructure requirements for the compute infrastructure to express an adaptation to the condition using a first LLM, wherein the infrastructure requirements are in human language;
modify an infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using one or more additional LLMs;
convert the modified infrastructure topology into an executable deployment specification for evolved compute infrastructure that conforms to the revised infrastructure requirements; and
execute the deployment specification to install the evolved compute infrastructure on the target computing system to cause the parameter to cease to detect the condition.

### Claim 15 (depends on 1, 14)

The computing system of claim 14, wherein the computer-executable instructions, when executed by at least the processor, further cause the computing system to perform the steps of claim 1 in a loop that iterates each time the condition is detected until an instruction to terminate the monitoring is received.

### Claim 16 (depends on 14)

The computing system of claim 14, wherein the computer-executable instructions, when executed by at least the processor, further cause the computing system to:
detect the condition because the compute infrastructure is overutilized or underutilized based on the parameter, wherein the parameter is a level of usage of the compute infrastructure; and
use the first LLM to revise the infrastructure requirements to adjust compute resources of the compute infrastructure.

### Claim 17 (depends on 14)

The computing system of claim 14, wherein the computer-executable instructions, when executed by at least the processor, further cause the computing system to:
detect the condition because an update to a component of the compute infrastructure has become available; and
use the first LLM to revise the infrastructure requirements to specify the component of the compute infrastructure to be configured according to the update.

### Claim 18 (depends on 14)

The computing system of claim 14, wherein the computer-executable instructions, when executed by at least the processor, further cause the computing system to:
prior to revising the infrastructure requirements, generate an initial logical infrastructure topology from the infrastructure requirements using a logical infrastructure topology generation LLM that is configured to generate graphs of logical infrastructure from the infrastructure requirements; and
generate the infrastructure topology using a physical infrastructure topology generation LLM that is configured to generate graphs of physical infrastructure from graphs of logical infrastructure and the infrastructure requirements, wherein the infrastructure topology is a physical infrastructure topology.

### Claim 19 (depends on 14)

The computing system of claim 14, wherein the computer-executable instructions to modify the infrastructure topology for the compute infrastructure to conform to the revised infrastructure requirements using the one or more additional LLMs, when executed by at least the processor, further cause the computing system to:
translate the revised infrastructure requirements into a current logical infrastructure topology using one of the additional LLMs that is trained to generate graphs of logical infrastructure from the infrastructure requirements and a prior logical infrastructure topology; and
translate the revised infrastructure requirements and the current logical infrastructure topology into a physical infrastructure topology using one of the additional LLMs that is trained to generate graphs of physical infrastructure from infrastructure requirements, the current logical infrastructure topology, and a prior physical infrastructure topology, wherein the modified infrastructure topology is a current physical infrastructure topology.

### Claim 20 (depends on 14)

The computing system of claim 14, wherein the executable deployment specification is written in YAML code or HCL code.
