Muhammad Maaz Aleem
Undergraduate Computer Science researcher and full-stack software engineer based in Islamabad, Pakistan. Rebuilding machine learning from first principles to solve critical bottlenecks in clinical imaging and biomedical text mining.
Engineering Depth & Foundations
My engineering foundation originated in building production web applications, distributed backend services, and automated data ingestion pipelines. Developing with TypeScript, Node.js, and MongoDB instilled a deep commitment to fault tolerance, type correctness, and systems that withstand irregular inputs.
However, the problems that hold my deepest curiosity aren't just about building faster storefronts or slicker interfaces — they reside at the intersection of computational modeling and human health. When artificial intelligence meets healthcare, the cost of a false negative isn't a lost cart conversion; it's a missed early diagnosis.
Rather than treating AI as an opaque black box accessed via third-party APIs, I am rebuilding my machine learning foundations from the mathematical roots up — writing autograd engines, backpropagation, and loss functions in pure NumPy and PyTorch before orchestrating complex transformer models and vision architectures.
Curriculum & Technical Roadmap
Andrew Ng's Machine Learning Specialization
Supervised learning, regularization, logistic regression, decision trees, and unrolled gradient descent.
PyTorch Foundations & Custom Notebooks
Rebuilding convolutional blocks, residual connections, attention matrices, and custom training loops.
fast.ai & Practical Deep Learning
State-of-the-art vision models, progressive resizing, learning rate finders, and medical image augmentation.
Hugging Face & Clinical NLP
BioBERT, PubMedBERT, sequence classification, and biomedical entity extraction pipelines.
LoRA & QLoRA Fine-Tuning
Parameter-efficient adaptation of open foundation weights on specialized clinical literature datasets.
Explore the four pillars of my current research
Detailed breakdown of deep learning, computer vision for screening, clinical NLP, and large pipelines.