# US 2026/0099734 A1 · System for automated data analysis and decision-making for complex product configuration

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

- Number: US 2026/0099734 A1
- Document ID: US20260099734A1
- Status: Published application
- Kind: A1
- Application number: 19/351,580
- Filed: 2025-10-07
- Published: 2026-04-09
- Earliest claimed: 2024-10-08
- Assignees: Vertiv Corporation
- Applicants: Vertiv Corporation
- Inventors: Paren Arvindbhai MAKWANA; Girish KADAM
- Web page: https://patentgrep.com/patent/US20260099734A1

## CPC

- G06N 5/04: Inference or reasoning models
- G06N 20/00: Machine learning
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 3/0985: Hyperparameter optimisation; Meta-learning; Learning-to-learn
- G06Q 30/0621: {by configuring or customising goods or services}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

The present invention relates to an AI-driven product configurator that pre-processes data collected from different sources, extracts relevant features from pre-processed data, infers results through usage of ML models, and verifies inferred results. Further, ML models are deployed on an edge device where computing resources are dynamically allocated. Detailed analytical reports include visualizations, summaries, and actionable insights are generated, and alerts or notifications are triggered based on predefined thresholds, detected anomalies, or critical conditions identified in the data.

## Claims

### Claim 1

A system for automated data analysis and decision-making, the system comprising:
an identifying module configured to identify a requirement based on a user input or pre-determined objectives;
a data collection module, operatively connected to the identifying module, configured to gather and compile data relevant to the identified requirement from one or more data sources;
a data pre-processing module, communicatively linked to the data collection module, configured to clean, normalize, and structure the data for further processing;
an inference module, operatively connected to the data pre-processing module, comprising one or more machine learning models, and configured to generate an output including one or more of insights, predictions, and classifications based on pre-processed data;
a feature extraction module, integrated with the inference module, configured to extract relevant features from the pre-processed data for enhanced processing and accurate decision-making;
a verification module, operatively linked to the inference module, configured to evaluate the output of the inference module to determine whether the output meets a predefined acceptability criterion;
a model deployment module, communicatively connected to the verification module, configured to deploy the one or more machine learning models in a real-time or offline environment based on verified results obtained from the verification module;
a model serving module, operatively linked to the model deployment module, configured to host deployed machine learning models for receiving a real-time input and providing corresponding predictions;
a feedback module, integrated with the model serving module, configured to collect real-time data and system performance metrics, and relay the real-time data and system performance metrics for updating the one or more machine learning models;
an edge deployment module, communicatively connected to the model deployment module, configured to deploy the one or more machine learning models on an edge device present near the one or more data sources, for optimized processing;
a hardware allocation module, integrated with the edge deployment module, configured to allocate computing resources including CPU, GPU, and RAM to support efficient model inference and execution;
a report generation module, operatively connected to the inference module, configured to generate reports based on the output of the inference module; and
an alert module, communicatively linked to the inference module and the report generation module, configured to trigger alerts based on the predictions.

### Claim 2 (depends on 1)

The system of claim 1, wherein the feedback module is further configured to update the one or more machine learning models based on real-time operational data to enhance model accuracy and performance.

### Claim 3 (depends on 1)

The system of claim 1, wherein the feedback module further comprises an explainability sub-module, utilizing SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Module-agnostic Explanations), to generate interpretable insights into the machine learning model's decision-making process.

### Claim 4 (depends on 1)

The system of claim 1, wherein the edge deployment module is further configured to optimize model performance by enabling low-latency inference and decision-making at the data source, reducing the need for centralized processing.

### Claim 5 (depends on 1)

The system of claim 1, wherein the edge deployment module implements a decentralized learning paradigm, enabling each edge device to perform localized training on data captured within its vicinity and contribute to a federated global model, ensuring privacy-preserving learning and reducing network congestion.

### Claim 6 (depends on 1)

The system of claim 1, wherein the verification module determines whether the output of the inference module meets predefined thresholds for accuracy.

### Claim 7 (depends on 1)

The system of claim 1, wherein the verification module further integrates a continuous integration/continuous deployment (CI/CD) pipeline to automatically validate and redeploy updated models based on new data inputs and system performance metrics.

### Claim 8 (depends on 1)

The system of claim 1, wherein the hardware allocation module dynamically adjusts resource allocation based on complexity of the data processed and demands of the inference module to maintain efficient system operation.

### Claim 9 (depends on 1)

The system of claim 1, wherein the hardware allocation module is further configured to employ memory-efficient inference techniques to optimize model performance in low-bandwidth environments.

### Claim 10 (depends on 1)

The system of claim 1, wherein the report generation module is configured to produce detailed analytics reports including visualizations, and transmit the analytics reports to a user interface for further action.

### Claim 11 (depends on 1)

The system of claim 1, wherein the report generation module further includes real-time data visualization capabilities, utilizing advanced graphical rendering techniques to provide interactive dashboards and reports that update in real time as new data is processed by the system.

### Claim 12 (depends on 1)

The system of claim 1, wherein the alert module is further configured to generate automated alerts in response to predefined conditions detected by the inference module.

### Claim 13 (depends on 1)

The system of claim 1, wherein the alert module utilizes an adaptive thresholding mechanism that dynamically adjusts alert thresholds based on contextual data patterns.

### Claim 14 (depends on 1)

The system of claim 1, wherein the feature extraction module is further configured to apply one or more of manual and automated feature selection techniques, to ensure the robustness and contextual relevance of extracted features in a presence of evolving data patterns.

### Claim 15 (depends on 1)

The system of claim 1, wherein the inference module implements an adaptive neural architecture search (NAS) framework capable of dynamically generating optimal model architectures, based on real-time evaluation of data complexity.

### Claim 16

A method of implementing automated data analysis and decision-making, comprising the steps of:
identifying a requirement or task through an identifying module, wherein the requirement is dynamically received from user input or system-generated objective;
collecting data from one or more data sources using a data collection module, wherein the one or more data sources include sensor networks, cloud storage, and edge devices, and the data is synchronized in real-time to ensure consistency;
pre-processing the data via a data pre-processing module, wherein the pre-processing involves cleaning, normalizing, reducing noise, and augmenting data for further processing;
extracting relevant features from pre-processed data using a feature extraction module;
inferring results through usage of one or more machine learning models by an inference module, wherein the inference includes one or more of generating predictions, classifications, and decisions using deep learning models and ensemble learning approaches, and wherein the model selection is dynamically adjusted based on input data characteristics;
verifying an output of the inference module using a verification module, wherein the verifying includes one or more of cross-validation, statistical analysis, and uncertainty quantification to determine whether the output meets a predefined acceptability criterion;
deploying the one or more machine learning models using a model deployment module, based on verified results obtained from the verification module;
serving the model through a model serving module, wherein real-time model requests are handled through a load-balancing mechanism and computational resources are dynamically scaled based on input complexity and system demand;
generating a feedback loop through a feedback module, wherein real-time system outputs and performance metrics are continuously monitored and used to update the one or more machine learning models through reinforcement learning or continuous model updates;
deploying the one or more machine learning models on an edge device via an edge deployment module, wherein the one or more machine learning model performs localized inference at the data source and can retrain locally on new data, reducing the need for centralized communication;
allocating computing resources through a hardware allocation module, wherein the computing resources including one or more of CPU, GPU, and RAM are dynamically managed across distributed environments to optimize performance during model inference;
generating reports using a report generation module, wherein the report generation module creates detailed analytical reports including visualizations, summaries, and actionable insights based on the output of the inference module; and
triggering alerts or notifications through an alert module, based on predefined thresholds, detected anomalies, or critical conditions identified in the data.

### Claim 17 (depends on 16)

The method of claim 16, wherein the step of pre-processing the data further includes applying context-aware normalization techniques that adapt to the type and variability of the data to optimize pre-processing performance.

### Claim 18 (depends on 16)

The method of claim 16, wherein the step of verifying the output of the inference module includes integrating a continuous integration and continuous deployment (CI/CD) pipeline, enabling automatic validation and redeployment of updated models based on system performance metrics and data changes.

### Claim 19 (depends on 16)

The method of claim 16, wherein the step of deploying the one or more machine learning models on the edge devices includes implementing a decentralized learning framework, wherein each edge device performs local model training on the data collected within its region and contributes to development of a federated global model, maintaining data privacy and reducing latency.

### Claim 20 (depends on 16)

The method of claim 16, wherein the method step of allocating computing resources includes employing a memory-efficient inference technique including zero-shot learning or federated learning, to optimize performance in low-bandwidth environments during edge deployments.
