Token-space search and decoding
I study language models through token-space structure, decoding dynamics, and controllability. The focus is reasoning diagnostics, decoding behavior, topology-aware analysis, and token-level interpretability.
/research
I study reasoning, reliability, and intelligence in AI systems.
My work spans token-space search in language models, validation layers for agents, and formal foundations for cognitive intelligence.
Research programs
Agent reliability and governance
I study validation layers, safety constraints, auditability, and governance mechanisms for agentic systems. This includes reliable execution, verification layers, provenance, policy enforcement, and agent safety infrastructure. NjiraAI appears here as public-facing governance infrastructure.
Foundations of intelligence
I study formal foundations for intelligence, cognition, and artificial minds. This line includes interface-plane intelligence measurement and recurrent state-transition models of cognition.
Selected public research
RUINES / interface-plane intelligence measurement
Define intelligence on an agent-environment interface plane using relevance partitions and regulation of predictive uncertainty.
AbstractWoI / Wheel of Intelligence
Model cognition as a recurrent state-transition loop over a shared state contract with seven core operators and governance patterns.
AbstractRUINES abstract
Intelligence is defined on an interface plane between agent and environment. Measurement is expressed through relevance partitions and regulation of predictive uncertainty.
WoI abstract
Cognition is modeled as a recurrent state-transition loop over a shared state contract, with seven core operators and governance patterns.
Research directions
topology-aware decoding
Structural analysis of decoding trajectories to improve controllability and interpretability.
agent governance
Validation and policy frameworks for safe, auditable, and reliable agent behavior.
geometric logic
Formal representations that connect geometric structure and reasoning processes.
embodied intelligence
Cognitive architectures for artificial minds, including links to humanoid cognitive systems.
Collaboration
I am open to collaboration with labs, co-authors, and technical research partners working on reasoning diagnostics, agent validation and governance, intelligence measurement, and cognitive architectures.
Audience
Labs, co-authors, technical research partners
Focus areas
Reasoning diagnostics, validation layers, intelligence measurement
Mode
Joint reading, shared protocols, and collaborative study design
Public research agenda, 2026.