Research & Innovation at MrTechLogix
Exploring Intelligent Systems at the Intersection of AI, Multimodal Learning and Decision Support
Research at MrTechLogix, led by CEO Khawar Maqsood, focuses on developing intelligent systems capable of learning from multiple information sources and supporting complex decision-making. This research discipline also informs practical technology solutions.
PhD Research
Pareto-Optimized Multimodal Reinforcement Learning for Radiological Decision Support Using MIMIC-CXR Chest Images and Clinical Reports
Research Pipeline
Medical Images + Clinical Reports → Multimodal Representation → Reinforcement Learning → Pareto Optimization → Decision Support
This research explores how an AI system can learn from both medical images and clinical text simultaneously. By combining computer vision (for processing chest X-ray images) and natural language processing (for interpreting clinical reports), the system develops a unified multimodal representation. A reinforcement learning agent then uses this representation to support complex radiological decision tasks. Pareto optimisation is applied to balance multiple competing objectives in the learning process — moving beyond single-metric optimisation towards more balanced and clinically meaningful outcomes. This is research-level work; no clinical deployment or medical claims are made.
Research Technologies
Core technologies powering this research.
Chest X-ray images
Clinical reports
Computer Vision
Natural Language Processing
Multimodal Learning
Reinforcement Learning
PPO
Pareto Optimization
NSGA-II
ResNet / Vision Transformer
BioClinicalBERT
PyTorch
Hugging Face
MONAI
Research Interests
Artificial Intelligence
Developing intelligent systems that can reason, learn and adapt across complex domains.
Machine Learning
Designing and evaluating supervised, unsupervised and semi-supervised learning models.
Reinforcement Learning
Training agents to make sequential decisions through interaction with dynamic environments.
Multimodal AI
Combining information from multiple modalities — text, images, structured data — into unified learning pipelines.
Computer Vision
Processing and understanding visual data including medical images using deep learning architectures.
Natural Language Processing
Extracting meaning, structure and knowledge from clinical and general-domain text.
Medical AI
Applying AI and machine learning techniques to support medical image analysis and clinical information processing at a research level.
Continual Learning
Building models capable of learning new tasks over time without forgetting previously acquired knowledge.
Federated Learning
Enabling model training across distributed data sources without centralising sensitive data.
Research & Technology Stack
Deep Learning
- ✓ PyTorch
- ✓ TensorFlow
NLP
- ✓ Hugging Face Transformers
- ✓ BioClinicalBERT
Computer Vision
- ✓ ResNet
- ✓ Vision Transformers
- ✓ MONAI
Machine Learning
- ✓ Scikit-learn
- ✓ Imbalanced-learn
Data
- ✓ Pandas
- ✓ NumPy
Visualization
- ✓ Matplotlib
- ✓ Power BI
Development
- ✓ Python
- ✓ WordPress
- ✓ Elementor
- ✓ HTML
- ✓ CSS
- ✓ JavaScript
- ✓ PHP
Research Workflow
- ✓ Git
- ✓ GitHub
- ✓ DVC
Explore the Research
Discover published work, methodology and the researcher behind MrTechLogix.