I am an independent researcher and final-year cybersecurity undergrad working at the intersection of machine learning, AI safety and adaptive systems. My research focuses on how machine learning systems behave under long-term adaptation, distributional shifts, and changing optimization dynamics. I study the statistical and structural properties of adaptive learning systems, investigating how architectural and training design choices influence robustness, generalization, controllability, and behavioral recoverability.

My recent work introduces Reversible Neural Adaptation (RLAE), a structural paradigm that separates behavioral learning from model identity, enabling deterministic rollback and addressing irreversible behavioral drift in weight-based adaptation. This work formalizes concepts such as structural irreversibility, recoverability, and behavioral divergence, supported by empirical evaluation across adaptation regimes and model scales. In parallel, I am developing a strong foundation in statistical machine learning and applied mathematics—particularly probability theory, optimization, and information theory—and applying these concepts to my research on latent behavioral structures in low-dimensional statistical manifolds for autonomous driving systems. My work explores how learning dynamics, manifold geometry, and distributional shifts shape behavioral adaptation, stability, interpretability, and robustness in autonomous world models and decision-making driving agents.

My research interests include statistical and robust machine learning, adversarial machine learning, structural robustness of adaptive models, AI safety and controllability, and runtime-governed modular learning systems. My goal is to contribute to the design of trustworthy machine learning systems that remain stable, interpretable, and controllable under continuous adaptation.

• Engineering runtime safety systems in Rust — kill-switches, isolation layers, fault boundaries
• Designing distributed agent runtimes in Go/K8s — CRDs, orchestration, gossip protocols
• Adversarial stress-testing MARL agents and behavioral LoRA modules
• Researching RLAE / reversible learning systems
• Papers:
  • On the Structural Limitations of Weight-Based Neural Adaptation... — arXiv
  • Temporal & Adaptive Dynamics of Latent Behavioral... — in progress

• Machine Learning (adaptive & robust systems)
• Systems & Distributed Architectures
• AI Safety & Statistical ML
• Cyber-Physical & Autonomous Systems

Aiming toward MSc → PhD focused on ML Research & AI Safety, Systems & Adaptive Intelligence.

If my research or experiments contribute to your projects or spark ideas, you can support my work here:

"We are merely vessels of a greater curiosity"