Filed 2024–2025, with systems deployed in Fortune-100 logistics at production scale — reported at ~200 ms latency across 216 languages with substantial GPU-cost reduction.
Patent list
WO/2025/141391
AI system for randomized dataset creation and evaluation of emotional-state detection in language models via multi-source aggregation and normalization.
WO/2025/233926
Hallucination-mitigation system for generative AI using neural verification, semantic-consistency scoring, and feedback correction loops.
WO/2025/233925
Multi-computer neural-architecture system for content generation using generator–discriminator frameworks and adaptive learning.
US 18/927,834
Neural-translation system using encoder–decoder architecture with domain adaptation, feedback learning, and contextual optimization.
Context
These patents emerged from work as a data scientist at Sunya Gen AI (2021–2022), where I co-developed agentic AI architectures for autonomous decision-making. The systems address a core enterprise problem: industrial data is heterogeneous, multilingual, and full of inconsistencies. Large language models can harmonise it — but only if their tendency to hallucinate and mis-translate can be controlled.