Academic publications and patent filings by M. Mostagir Bhuiyan, covering CPU architecture, AI systems, and climate science.
Patent
PATENT PENDING / USPTO
Micro-Containerized CPU Architecture for Efficient AI Workloads
Application No. 19/262,056 / Filed July 2025
M. Mostagir Bhuiyan holds USPTO Patent Application 19/262,056 for a micro-containerized CPU architecture designed to run AI workloads more efficiently. The invention partitions individual CPU cores into isolated, lightweight processing units called "micro containers." Each micro container operates with its own execution context, enabling fine-grained parallel processing that approaches GPU-level throughput for specific AI tasks.
The patent was originally filed as Provisional Application No. 63/794,191 in April 2025, then converted to a full US Utility Patent application in July 2025. The architecture targets inference workloads where GPU availability is constrained or cost prohibitive, offering a CPU-native alternative for deploying machine learning models at scale.
Inventor: M. Mostagir Bhuiyan
Publications
TechRxiv
Hypothetical Framework for CPU Micro Containerization: Bridging the Performance Gap with GPUs in AI
M Mostagir Bhuiyan
M. Mostagir Bhuiyan authored this paper proposing a framework for CPU micro-containerization. The paper introduces a method to maximize Central Processing Unit throughput by dissecting CPU cores into isolated, efficient processing units called "micro containers." These micro containers simulate GPU capabilities for parallel processing, targeting the performance gap between CPUs and GPUs in AI workloads.
The framework theoretically partitions each physical core into multiple execution contexts with independent memory spaces, register files, and scheduling queues. This enables a single CPU to handle workloads that traditionally require GPU hardware, with particular applicability to inference tasks in resource-constrained environments.
The Illusion of Boundless AI: Analyzing Limitations and Ethical Concerns
M Mostagir Bhuiyan
M. Mostagir Bhuiyan authored this paper arguing against the rhetoric of unlimited AI potential. The paper examines the constraints currently facing AI and machine learning models, focusing on the finitude of training data, diminishing returns from scaling, and the gap between benchmark performance and real-world reliability.
The analysis covers data quality and availability ceilings, computational cost scaling laws, the challenge of generalization outside training distributions, and ethical concerns around deployment without adequate safeguards. The paper calls for more measured expectations and engineering-grounded evaluation of AI capabilities.
Retrieval-Native Language Models: Integrating Parametric and Vector Memory with Bayesian Attention
M Mostagir Bhuiyan
M. Mostagir Bhuiyan authored this paper proposing Retrieval-Native Language Models (RLLMs), a new paradigm that treats vector-based memory as a first-class component of the model architecture. Unlike retrieval-augmented generation (RAG), which bolts external retrieval onto existing models, RLLMs integrate three channels of knowledge directly: parametric weights, vector memory, and live retrieval.
The architecture uses Bayesian attention mechanisms to dynamically weight contributions from each knowledge channel based on query context and confidence estimation. This enables the model to seamlessly blend learned knowledge with retrieved facts, reducing hallucination and improving factual grounding without the latency and pipeline complexity of traditional RAG systems.
Technological Adaptation Outpaces Climate Impacts on Aviation: Evidence from Three Decades of Warming
M Mostagir Bhuiyan & Rifa Rafia
M. Mostagir Bhuiyan co-authored this paper analyzing the relationship between rising global temperatures and aviation efficiency over a thirty-year period. The study presents quantitative evidence that technological improvements in engine design, aerodynamics, and operational procedures have outpaced the negative effects of climate change on flight performance.
The research examines fuel consumption data, takeoff performance metrics, and atmospheric density changes across multiple decades. Findings indicate that while higher temperatures reduce air density and increase required takeoff distances, advances in aircraft technology have more than compensated for these effects, resulting in net efficiency gains industry-wide.