US 12,470,421 B2
Data analytics platform for stateful, temporally-augmented observability, explainability and augmentation in web-based interactions and other user media
Thomas N. Blair, Irvine, CA (US); Alexey Goder, San Jose, CA (US); Joerg Rings, Vancouver, WA (US); Joshua Peter Francis Yoder, Rochester, NY (US); Spyros J. Lazaris, Los Angeles, CA (US); and Gregory Burlet, Edmonton (CA)
Assigned to AGBLOX, INC., Irvine, CA (US)
Filed by AGBLOX, INC., Irvine, CA (US)
Filed on Oct. 11, 2023, as Appl. No. 18/379,118.
Claims priority of provisional application 63/415,266, filed on Oct. 11, 2022.
Prior Publication US 2024/0121125 A1, Apr. 11, 2024
Int. Cl. H04L 12/18 (2006.01); G06F 16/34 (2025.01); G06F 40/58 (2020.01); G06V 20/20 (2022.01); G06V 20/40 (2022.01)
CPC H04L 12/1831 (2013.01) [G06F 16/345 (2019.01); G06F 40/58 (2020.01); G06V 20/47 (2022.01)] 27 Claims
OG exemplary drawing
 
1. A method, comprising:
ingesting input data comprised of one or more files representing at least one business process workflow;
modeling the input data to analyze a content of the one or more files in one or more machine learning models and execute the one or more business process workflows, the one or more machine learning models configured to:
extract the content in the one or more files, wherein one or more embeddings are generated from the one or more files,
create at least one embeddings vector and applying a vector cosine similarity to the at least one embeddings vector, and
apply one or more knowledge graphs to the content, the one or more knowledge graphs including information from a plurality of external data sources that is integrated with the content one or more files to develop an explainability layer that annotates the content by constructing a structured representation of the content using data points extracted from the one or more files, and associating the data points with particular, domain-specific information inferred from the external data sources that is relative to the content;
developing one or more domain-specific neural networks from the content, the domain-specific neural networks configured to analyze the structured representation of the content and the domain-specific information inferred from the external data sources;
augmenting the one or more domain-specific neural networks from the explainability layer over time to create temporal annotations of user-identified information from the one or more files to execute the at least one business process workflow; and
autonomously extracting the user-identified information from the content in response to one or more prompts of a large language model, wherein the one or more prompts are configured from the content and based on the at least one business process workflow defined within an artificial intelligence-based agent and curated to achieve user-defined outcomes in the at least one business process workflow, and
executing the at least one business process workflow, the artificial intelligence-based agent automatically assembled to execute the at least one business process workflow based on the user-identified information, wherein the explainability layer enables a traceability in the one or more workflows between the one or more files and the user-defined outcomes, wherein the artificial intelligence-based agent maintains a stateful execution of the at least one business workflow outside of conversational interactions.