MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202621085063 A) filed by Priyanka V. Deshmukh; Amit Kanskar; Amir Singh; Vikas Singhai; Sonali Namdeo; Tanmay Tripathi; Priya Pathak; Deepak Singh; and Rizwan Khan on July 10, 2026, for Hybrid Multimodal Ai Framework For Hidden Talent Extraction Using Behavioural, Linguistic, And Coding Pattern Analysis With Explainable Outputs.
Inventors include Amit Kanskar; Amir Singh; Vikas Singhai; Sonali Namdeo; Tanmay Tripathi; Priya Pathak; Deepak Singh; and Rizwan Khan.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: The present invention relates to a hybrid multimodal artificial intelligence framework and method for automated hidden talent extraction through integrated analysis of behavioural, linguistic, and coding patterns with explainable decision generation. The invention acquires heterogeneous data from multiple sources, including user interaction logs, textual communications, programming artefacts, assessment records, and digital activity traces, and processes the collected information through behavioural analytics, natural language processing, and static and dynamic coding analysis to extract meaningful features representing engagement, cognitive abilities, communication skills, problem-solving aptitude, algorithmic thinking, coding efficiency, and programming behaviour. The extracted multimodal features undergo preprocessing, normalization, feature selection, and multimodal feature fusion using a hybrid artificial intelligence engine comprising machine learning algorithms, deep learning architectures, transformer-based language models, sequence modelling networks, and ensemble learning techniques to generate comprehensive talent profiles and predict latent competencies that are not readily identifiable through conventional assessment systems. An Explainable Artificial Intelligence (XAI) module provides interpretable outputs by determining the contribution of each data modality, generating feature importance scores, confidence measures, visual explanations, and human-readable decision justifications, thereby improving transparency, fairness, and trust in automated talent prediction. The framework further incorporates adaptive learning mechanisms for continuous model refinement using newly acquired behavioural and performance data, enabling progressive improvement in prediction accuracy and robustness. The invention facilitates early identification of hidden talents, personalized competency mapping, intelligent career guidance, recruitment support, educational assessment, and workforce development while providing a scalable, reliable, and ethically responsible artificial intelligence solution that significantly outperforms conventional unimodal talent assessment approaches in terms of prediction accuracy, explainability, and decision reliability.
Disclaimer: Curated by HT Syndication.
