MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202621088532 A) filed by Praavega Ai Private Limited on July 20, 2026, for System And Method For Dynamic Substitution Of Base Machine-Learning Models In A Model-Serving Runtime.
Inventors include Anshoo Gaur; and Lakshmanaraj Sankaralingam.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: ABSTRACT SYSTEM AND METHOD FOR DYNAMIC SUBSTITUTION OF BASE MACHINE-LEARNING MODELS IN A MODEL-SERVING RUNTIME The present disclosure relates to a system (100) and a method (1000) for substituting a first base machine-learning model with a second base machine-learning model in a model-serving runtime. The method (1000) includes receiving and cryptographically verifying a machine-readable base descriptor of the second base machine-learning model, executing a conformance workload, generating a cryptographically bound conformance report, and determining a compatibility parity verdict. The method (1000) further includes obtaining adapter provenance records and performing bind-time validation based on a target base-model generation and a target tokenizer fingerprint. Upon successful validation, a signed composition attestation is generated identifying a validated model-serving stack comprising the second base machine- learning model and one or more adapter artifacts. The validated model-serving stack is then activated in the model-serving runtime, and inference workloads previously serviceable by the first base machine-learning model are routed to the second base machine-learning model.
Disclaimer: Curated by HT Syndication.
