# US 2026/0094101 A1 · Machine learning framework that extracts actionable insights from disparate data sources > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2026/0094101 A1 - Document ID: US20260094101A1 - Status: Published application - Kind: A1 - Application number: 18/902,874 - Filed: 2024-09-30 - Published: 2026-04-02 - Earliest claimed: 2024-09-30 - Assignees: Intuit Inc. - Applicants: Intuit Inc. - Inventors: Siddharth JAIN; Sivashanker THIRUCHITTAMPALAM; Venkat Narayan VEDAM; Jonathan LIN - Web page: https://patentgrep.com/patent/US20260094101A1 ## CPC - G06Q 10/06375: {Prediction of business process outcome or impact based on a proposed change} - G06F 16/248: Presentation of query results - G06F 16/258: {Data format conversion from or to a database} - 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 - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Machine learning framework that extracts actionable insights from disparate data sources include performing operations. A merged feature set is generated by collecting diverse data to generate aggregated data in a unified dataset and performing an autoencoding of the aggregated data to generate compressed data in a low-dimensional latent space. The merged feature set is further generated by merging the first attribute and the second attribute to obtain a merged feature set for the low-dimensional latent space. The operations include selecting a response vector from the merged feature set in the low-dimensional latent space that is aligned with a user query, decoding the response vector into natural language text, executing a large language mode (LLM) to the natural language text and the user query to generate an actionable insight. ## Claims ### Claim 1 A method comprising: selecting a response vector from a merged feature set in a low-dimensional latent space that is aligned with a user query, wherein the merged feature set is generated by: collecting diverse data to generate aggregated data in a unified dataset, wherein the diverse data comprises first data from a first source and second data from a second source, the second source being disparate from the first source, the first data having a first attribute, and the second data having a second attribute, performing an autoencoding of the aggregated data to generate compressed data in the low-dimensional latent space, wherein the compressed data are representations of the aggregated data in the low-dimensional latent space, and merging the first attribute and the second attribute to obtain the merged feature set for the low-dimensional latent space; decoding the response vector into natural language text; executing a large language mode (LLM) to the natural language text and the user query to generate an actionable insight; and presenting the actionable insight as a response to the user query. ### Claim 2 (depends on 1) The method of claim 1, wherein selecting the response vector comprises: identifying a region of the low-dimensional latent space aligned with the user query; and selecting the response vector from the region of the low-dimensional latent space based on the merged feature set. ### Claim 3 (depends on 2) The method of claim 2, further comprising: selecting responsive compressed data in the region, the responsive compressed data being identified as similar to the user query based on the merged feature set. ### Claim 4 (depends on 3) The method of claim 3, further comprising: applying a continuous optimization algorithm to the user query and the low-dimensional latent space to locate the region. ### Claim 5 (depends on 3) The method of claim 3, wherein the responsive compressed data is within a threshold to the user query in the low-dimensional latent space. ### Claim 6 (depends on 5) The method of claim 5, wherein the responsive compressed data is semantically similar to the user query. ### Claim 7 (depends on 4) The method of claim 4, wherein the continuous optimization algorithm is one of: a gradient descent algorithm, an adaptive moment estimation algorithm, and a Bayesian optimization algorithm. ### Claim 8 (depends on 1) The method of claim 1, further comprising: concatenating the natural language text and the user query to form a concatenated string; and applying the LLM to the concatenated string. ### Claim 9 (depends on 1) The method of claim 1, wherein performing the autoencoding is by an autoencoder neural network. ### Claim 10 (depends on 1) The method of claim 1, further comprising: identifying a first format of the first data; and modifying the second data from a second format to the first format. ### Claim 11 (depends on 10) The method of claim 10, wherein the first format is one of a date format and a numerical value format. ### Claim 12 (depends on 1) The method of claim 1, wherein the first data is semi-structured data, collecting the diverse data comprises parsing the semi-structured data to generate structured data, and the aggregated data comprises the structured data. ### Claim 13 (depends on 12) The method of claim 12, further comprising: extracting nested structures and attributes from the first data using hierarchical markdown parsing. ### Claim 14 (depends on 12) The method of claim 12, further comprising: extracting nested structures and attributes from the first data using key-value parsing. ### Claim 15 (depends on 12) The method of claim 12, further comprising: extracting structure elements from metadata attributes associated with the first data. ### Claim 16 A system comprising: a server comprising a processor; a data repository in communication with the processor, and configured to store: diverse data comprising first data from a first source and second data from a second source, the second source being disparate from the first source, the first data having a first attribute, and the second data having a second attribute, compressed data in a low-dimensional latent space, and a merged feature set for the low-dimensional latent space, a large language mode (LLM), wherein the processor is programmed to apply the LLM to natural language text and a user query to generate an actionable insight; and a server controller executable by the processor to perform operations comprising: collecting the diverse data to generate aggregated data in a unified dataset, performing an autoencoding of the aggregated data to generate the compressed data, merging the first attribute and the second attribute to obtain the merged feature set, selecting a response vector from the merged feature set in the low-dimensional latent space that is aligned with the user query, wherein the merged feature set is generated by: collecting the diverse data to generate the aggregated data in the unified dataset, performing the autoencoding of the aggregated data to generate the compressed data in the low-dimensional latent space, wherein the compressed data are representations of the aggregated data in the low-dimensional latent space, and merging the first attribute and the second attribute to obtain the merged feature set for the low-dimensional latent space, decoding the response vector into the natural language text, executing the LLM on the natural language text and the user query to generate the actionable insight, and presenting the actionable insight as a response to the user query. ### Claim 17 (depends on 16) The system of claim 16, wherein the operations further comprise: selecting responsive compressed data in a region, the responsive compressed data being identified as similar to the user query based on the merged feature set. ### Claim 18 (depends on 17) The system of claim 17, wherein the responsive compressed data is within a threshold to the user query in the low-dimensional latent space. ### Claim 19 (depends on 16) The system of claim 16, wherein the operations further comprise: applying a continuous optimization algorithm to the user query and the low-dimensional latent space to locate the region. ### Claim 20 A non-transitory computer readable storage medium storing computer readable program code which, when executed by at least one processor, cause the at least one processor to perform operations comprising: selecting a response vector from a merged feature set in a low-dimensional latent space that is aligned with a user query, wherein the merged feature set is generated by: collecting diverse data to generate aggregated data in a unified dataset, wherein the diverse data comprises first data from a first source and second data from a second source, the second source being disparate from the first source, the first data having a first attribute, and the second data having a second attribute, performing an autoencoding of the aggregated data to generate compressed data in the low-dimensional latent space, wherein the compressed data are representations of the aggregated data in the low-dimensional latent space, and merging the first attribute and the second attribute to obtain the merged feature set for the low-dimensional latent space; decoding the response vector into natural language text; executing a large language mode (LLM) to the natural language text and the user query to generate an actionable insight; and presenting the actionable insight as a response to the user query.