Akshat Singh is a Robotics Engineer at Addverb, Noida, India, with an interdisciplinary background in robotics, automation, artificial intelligence, and natural language processing. He holds a Bachelor of Technology in Mechatronics, Robotics, and Automation Engineering from Symbiosis Institute of Technology, Pune, where he studied from 2022 to 2026. Before joining Addverb, Akshat worked as an AI Researcher Intern at Hosei University, Tokyo, Japan, focusing on the development of lightweight and efficient NLP and large language models. His research involved quantization, pruning, knowledge distillation, and knowledge graph optimization. During his internship, he contributed to reducing inference latency by approximately 30%, compressing LLM size by 40% while maintaining 95% accuracy, and developing an automated evaluation pipeline that reduced benchmarking time by 60%. Akshat's professional interests lie at the intersection of robotics, artificial intelligence, intelligent automation, and efficient machine learning systems. His international research experience and engineering background enable him to approach complex technological challenges from both research and practical perspectives. Currently, he is working as a Robotics Engineer at Addverb, where he contributes to the development and application of robotics and automation technologies for real-world industrial environments.  

MASG-AIA: A Deterministic, Validation-Gated Cyber-Cognitive Intelligence Architecture for Cross-Domain Threat Reasoning   Cyber threat intelligence systems typically model technical artifacts, endpoint events, social narratives, and human cognitive effects as separate analytical domains. This separation creates a multi-domain evasion problem: an adversary can distribute weak signals across infrastructure, operating-system telemetry, social information channels, and analyst cognition such that no single domain crosses a detection or confidence threshold. This paper presents MASG-AIA, a deterministic and validation-gated architecture for cyber-cognitive intelligence. The Multi-Domain Adversarial Signal Graph (MASG) is a typed property multigraph that jointly represents cyber entities, OSlevel observations, social-network narratives, cognitive-bias activations, and defensive countermeasures with temporal validity, provenance, and risk decomposition. The Autonomous Intelligence Architecture (AIA) adds deterministic engines for crossdomain chain discovery, narrative-cyber correlation, poisoning detection, graph uncertainty quantification, schema evolution, counterfactual reasoning, and autonomous hypothesis generation. Unlike runtime neural or prompt-based systems, MASG-AIA uses LLM assistance only during offline human-supervised curation; all runtime analysis is performed by auditable graph and statistical algorithms. We provide formal definitions, algorithms, a threat model, an operational workbench design, fixture-based tests, simulation-backed ablation and red-team experiments, a defensive response layer inspired by countermeasure knowledge graphs, and a validation gate protocol for moving from prototype evidence to public benchmark and production SOC claims. The evaluation currently demonstrates internal consistency, deterministic execution, and expected degradation under ablations, but it is deliberately claim-bounded: external public-dataset benchmarking and production SOC validation remain gates rather than asserted outcomes. The result is a submission-ready framework for reproducible research on cross-domain cybercognitive threat reasoning. Index Terms—Cyber threat intelligence, knowledge graphs, graph security, cognitive security, disinformation, adversarial reasoning, uncertainty quantification, cyber defense, provenance, validation gate