Curriculum Vitae
MD Azizul Hakim
AI Researcher
AI Researcher | Lead Author (Nature Portfolio & IEEE) | Biologically-Inspired Neural Architectures
Dhaka, Bangladesh
Research Summary
Independent researcher advancing the intersection of neuroscience and artificial intelligence through biologically-inspired computational frameworks. My work focuses on developing neural network architectures that embody biological principles of homeostasis, neuroplasticity, and adaptive repair to achieve robust, resilient, and biologically-faithful artificial intelligence systems. With publications in Nature Portfolio and IEEE conferences, I have demonstrated that integrating mechanisms from biological nervous systems—including calcium-dependent regulation, multi-scale temporal coordination, and self-repair capabilities—substantially enhances the stability and generalization of machine learning models across diverse domains including molecular property prediction, computer vision, and healthcare AI.
Publications
Journal Articles
Biologically inspired neural network layer with homeostatic regulation and adaptive repair mechanisms
2025
Scientific Reports (Nature Portfolio), Vol. 15, Article 9114
Authors: Hakim, M.A. & Alam, M.I.
DOI: 10.1038/s41598-025-09114-8
DOI: 10.1038/s41598-025-09114-8
- Introduced first implementation of homeostatic regulation mechanisms in artificial neural network layers
- Demonstrated superior performance on critical biomedical benchmarks including AIDS, HIV, COX2, and protein analysis datasets
- Established theoretical foundation for biologically-faithful AI architectures through adaptive repair mechanisms
- Validated self-repair capabilities under various perturbation scenarios, advancing robust AI systems
Conference Proceedings
A Novel Neuroplasticity-Based Approach for Enhanced Generalization in Computer Vision
2025
IEEE International Conference Proceedings
Authors: Hakim, M.A., Ifty, R.A., Delowar, K.E., Shuvo, S.A., Tasneem, Z., & Alam, M.I.
DOI: 10.1109/QPAIN66474.2025.11172203
DOI: 10.1109/QPAIN66474.2025.11172203
- Developed BioLogicalNeuron architecture incorporating calcium-dependent regulation and homeostatic control mechanisms
- Achieved 90.42% accuracy on CIFAR-10 (3.77% improvement over baseline), 99.43% on MNIST, and 93.27% on Fashion-MNIST
- Demonstrated that biological principles of neuroplasticity significantly enhance model generalization across diverse visual domains
- Established repair threshold methodology (0.8) maintaining stability metrics above 0.92 through adaptive interventions
NutriGuard: LLM-Driven Nutritional Assessment for Chronic Disease Prevention
2025
IEEE International Conference Proceedings
Authors: Hakim, M.A., Ifty, R.A., Delowar, K.E., Chowdhury, S.H., Rashid, I., & Shakib, M.
DOI: 10.1109/QPAIN66474.2025.11171750
DOI: 10.1109/QPAIN66474.2025.11171750
- Designed multimodal AI framework integrating OCR, deep learning, and fine-tuned LLMs for personalized nutrition management
- Implemented hybrid OCR pipeline combining PaddleOCR (English) and Surya (Bengali) with CNN-based fallback system
- Fine-tuned Llama-3.2 model on clinical guidelines to generate health condition-specific dietary recommendations
- Established multilingual, clinically-validated framework for dietary risk mitigation addressing chronic disease prevention
Neuroadaptive Intelligence: A Biologically-Inspired, Self-Repairing Neural Layer with Adaptive Learning for Molecular Graph Classification
2025
IEEE International Conference Proceedings
Author: Hakim, M.A. (sole author)
DOI: 10.1109/QPAIN66474.2025.11171860
DOI: 10.1109/QPAIN66474.2025.11171860
- Developed self-repairing neural layer architecture with adaptive learning mechanisms for molecular graph classification
- Demonstrated neuroadaptive intelligence principles in scientific computing applications
- Advanced biologically-inspired AI methodologies for drug discovery and molecular property prediction
- Sole-author publication demonstrating independent research capability
Research in Progress
Multi-Scale Temporal Homeostasis Enables Efficient and Robust Neural Networks
Manuscript in Preparation
Target Journal: Nature Machine Intelligence
Developing Multi-Scale Temporal Homeostasis (MSTH), the first cross-scale coordination framework integrating regulation across four temporal scales (ultra-fast: 5ms, fast: 2s, medium: 5min, slow: 1hr) for artificial neural networks.
- First implementation of hierarchical temporal coordination spanning milliseconds to hours in neural architectures
- Demonstrates consistent accuracy improvements and elimination of catastrophic failures across molecular, graph, and image domains
- Outperforms both single-scale bio-inspired models and established state-of-the-art methods
- Establishes cross-scale temporal coordination as fundamental principle for building resilient AI systems
Research Interests
- Biologically-Inspired Neural Architectures: Developing computational frameworks that incorporate homeostatic regulation, neuroplasticity, self-repair mechanisms, and multi-scale temporal dynamics inspired by biological nervous systems
- Robust and Resilient AI Systems: Addressing fundamental brittleness of artificial neural networks through bio-inspired stability mechanisms that maintain consistent performance under perturbations, distribution shifts, and adversarial conditions
- Multi-Scale Temporal Coordination: Investigating how coordination across multiple temporal scales enhances computational efficiency, stability, and recovery capabilities in neural systems
- Neuromorphic and Brain-Inspired Computing: Translating principles from neuroscience, systems biology, and evolutionary theory into computational frameworks that advance artificial intelligence capabilities
- AI for Healthcare and Drug Discovery: Applying biologically-principled AI architectures to biomedical domains including molecular property prediction, protein analysis, and personalized medicine
Honors & Recognition
Kaggle Notebooks Master
2024
Achieved Master tier on Kaggle through sustained creation of high-quality computational notebooks demonstrating expertise in data analysis, machine learning implementation, statistical modeling, and technical communication to global data science community.
2nd Runner-Up, National AI Agent Development Competition
2025
Intent Company & International Islamic University Chittagong
Secured 2nd Runner-Up position among 200+ competing teams in Bangladesh's premier AI agent development competition, demonstrating practical expertise in multi-agent systems, autonomous decision-making architectures, and real-time AI orchestration.
Selected Research Implementations
Biological Self-Healing Neural System
Open-source implementation of homeostatic regulation and adaptive repair mechanisms for artificial neural networks, published in Scientific Reports.
- Custom PyTorch layer architecture implementing biological feedback mechanisms for network stability
- Graph neural network applications for molecular property prediction on AIDS, HIV, and COX2 datasets
- Self-repair algorithms validated under various perturbation scenarios
- Modular design enabling integration into existing deep learning pipelines
PyTorch Graph Neural Networks Biologically-Inspired AI Molecular ML
FounderAI: Multi-Agent Executive Intelligence Platform
Production-grade multi-agent orchestration system providing AI-powered strategic capabilities through specialized executive agents (CMO, CPO, CTO, COO).
- Multi-agent coordination using CrewAI and LangGraph frameworks for complex decision-making workflows
- Model Context Protocol (MCP) implementation enabling seamless inter-agent communication
- Specialized agent personas with domain expertise in marketing, product, technology, and operations
- Integration of tool-calling capabilities for market research, financial modeling, and technical planning
LangChain LangGraph Multi-Agent Systems CrewAI
NutriGuard: Intelligent Food Assessment System
Multimodal AI framework for real-time nutritional assessment and personalized dietary recommendations based on health conditions.
- Hybrid OCR pipeline with multilingual support (English via PaddleOCR, Bengali via Surya)
- CNN-based food category detection with nutritional database cross-referencing
- Fine-tuned Llama-3.2 model for clinical guideline-based dietary analysis
- RAG system integrating medical literature for evidence-based recommendations
Computer Vision OCR LLM Fine-tuning Healthcare AI
Technical Expertise
Programming Languages
Python (Advanced) MATLAB SQL JavaScript
Machine Learning Frameworks
PyTorch TensorFlow Scikit-learn Hugging Face Transformers PyTorch Geometric
Research Specializations
Graph Neural Networks Computer Vision Natural Language Processing Molecular Machine Learning Biomedical AI Neuromorphic Computing
AI Engineering Tools
LangChain LangGraph CrewAI LlamaIndex RAG Systems Vector Databases
Development & Deployment
Git/GitHub Docker FastAPI MLflow AWS
Education
Diploma in Computer Science and Engineering
2021 - 2025 (Expected)
Bangladesh Sweden Polytechnic Institute, Dhaka, Bangladesh
Comprehensive program covering computer science fundamentals with specialized focus on artificial intelligence, machine learning, and software engineering. Conducted independent research in biologically-inspired neural network architectures, resulting in multiple peer-reviewed publications while pursuing diploma studies.
Relevant Coursework:
- Artificial Intelligence & Machine Learning
- Data Structures & Algorithms
- Database Management Systems
- Software Engineering & System Design
- Computer Networks & Operating Systems
Languages
Bengali
Native Proficiency
English
Professional Working Proficiency
Research publication, technical documentation, international collaboration
Professional Activities
- Open Source Contributions: Active contributor to biologically-inspired AI research with public implementations on GitHub
- Peer Review: Experienced in academic writing and research methodology aligned with top-tier journal standards
- Technical Communication: Kaggle Notebooks Master demonstrating ability to communicate complex concepts to diverse audiences
- Community Engagement: Participation in competitive AI challenges and national-level competitions