US 12,423,519 B1
Applying cascading machine learning models to command prompts
Maneet Sharma, New York, NY (US); Adar K. Danait, New York, NY (US); Flora P. Sah, New York, NY (US); Krishnendu Chatterjee, New York, NY (US); Rama Koteswara Rao Kandimalla, New York, NY (US); Pavithra Araleri Visweswariah, New York, NY (US); Djona Fegnem, New York, NY (US); Swati Mittal, New York, NY (US); Paul Joseph Zakharia, New York, NY (US); Paul Edward Hibbs, New York, NY (US); and James Michael Iacona, New York, NY (US)
Assigned to CITIBANK, N. A.
Filed by Citibank, N.A., New York, NY (US)
Filed on Jan. 10, 2025, as Appl. No. 19/017,003.
Application 19/017,003 is a continuation in part of application No. 18/582,425, filed on Feb. 20, 2024.
Int. Cl. G06F 40/20 (2020.01); G06F 21/57 (2013.01)
CPC G06F 40/20 (2020.01) [G06F 21/577 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system for applying cascading machine learning models to command prompts, the system comprising:
a storage device; and
one or more processors communicatively coupled to the storage device storing instructions thereon, that cause the one or more processors to:
receive a natural language query indicating a computing process to be performed;
generate, based on the natural language query, a command prompt for a first instance of a large language generative model, wherein the command prompt instructs the first instance of the large language generative model to output a plurality of activities for performing the computing process;
input, into the first instance of the large language generative model, the command prompt to cause the first instance of the large language generative model to output the plurality of activities associated with the computing process, wherein each activity of the plurality of activities comprises an activity-related natural language response, wherein the first instance of the large language generative model is trained to predict activities based on natural language queries, and wherein the large language generative model (i) extracts one or more items from inputs to the large language generative model, (ii) retrieves data from one or more data sources based on the one or more items, and (ii) generates outputs based on the data;
input, into a second instance of the large language generative model, a first activity-related natural language response associated with a first activity of the plurality of activities to cause the second instance of the large language generative model to output a plurality of vulnerabilities associated with the first activity, wherein each vulnerability of the plurality of vulnerabilities comprises a vulnerability-related natural language response, and wherein the second instance of the large language generative model is trained to predict vulnerabilities based on activities;
input, into a third instance of the large language generative model, a first vulnerability-related natural language response associated with a first vulnerability of the plurality of vulnerabilities to cause the third instance of the large language generative model to output one or more control tools for addressing the first vulnerability, wherein each control tool of the one or more control tools comprises a control-related natural language response, and wherein the third instance of the large language generative model is trained to identify, for vulnerabilities, control tools of a plurality of available control tools;
input, into a fourth instance of the large language generative model, a first control-related natural language response corresponding to a first control tool of the one or more control tools to cause the fourth instance of the large language generative model to output one or more monitoring tools for monitoring the first control tool, wherein each monitoring tool of the one or more monitoring tools comprises a monitoring-related natural language response, and wherein the fourth instance of the large language generative model is trained to identify monitoring tools for control tools; and
in response to the natural language query, generate for display the first activity-related natural language response, the first vulnerability-related natural language response, the first control-related natural language response, and a first monitoring-related natural language response corresponding to a first monitoring tool of the one or more monitoring tools.