Muhammad Maaz Aleem
Undergraduate Computer Science Researcher · Clinical AI & Domain Shift in Medicine
Research Statement & Focus
Investigating Domain Shift and Out-of-Distribution (OOD) generalization in clinical machine learning. Deep neural networks trained on medical scans from a single hospital scanner frequently experience severe diagnostic performance degradation when evaluated on external clinical sites due to acquisition artifacts, scanner vendor variability (e.g., Siemens vs. GE), demographic shifts, and spurious shortcut features. My research focuses on developing invariant feature representations, test-time adaptation strategies, and uncertainty quantification in PyTorch to ensure diagnostic algorithms maintain robust clinical reliability across real-world healthcare networks.
Education
Air University, Islamabad
Bachelor of Science in Computer Science (BSCS)
Key Coursework: Data Structures & Algorithms, Linear Algebra, Probability & Statistics, Object-Oriented Programming, Discrete Mathematics, Software Engineering.
Experience Record
Machine Learning Intern · FlyRank
Internship- Architected high-throughput tabular data processing and transformation pipelines in Python & Pandas operating on a data warehouse containing tens of millions of records.
- Addressed memory consumption bottlenecks and eliminated out-of-memory (OOM) failures by designing chunked tensor ingestion routines.
- Conducted rigorous exploratory data analysis (EDA) and automated data cleansing workflows in Jupyter to detect schema drifts and distribution outliers.
Full-Stack Software Engineering Intern · DevelopersHub Corporation
Internship- Constructed resilient full-stack e-commerce architecture utilizing TypeScript, React, Node.js, Express, and MongoDB.
- Engineered secure JWT-authenticated API endpoints, schema indexing, and role-based permissions handling concurrent checkout workflows.
Freelance Full-Stack & Automation Engineer · Fiverr (@maaz_aleem64)
Ongoing- Delivered custom web applications, administrative systems, and dynamic web scraping robots for international clients.
- Implemented robust JavaScript rendering workarounds, automated rate-limiting protections, and structured dataset parsers.
Selected Research & Engineering Projects
FlyRank ML — Big Data & Warehouse Pipeline
GitHub RepositoryScalable data transformation engine capable of processing tens of millions of records. Built with Python, Pandas, NumPy, and Jupyter.
Coffee Shop — MERN Commercial Platform
GitHub RepositoryFull-stack commerce storefront featuring JWT authentication, reactive cart state management, and optimized MongoDB document indexing.
PyTorch Foundations & Domain Shift Experiments (In Progress)
Academic NotebooksRebuilding deep neural network autograd, backpropagation, and loss functions from scratch, benchmarking domain generalization losses on imbalanced diagnostic imaging datasets.
Technical Skills Matrix
Independent Curriculum & Certifications
- Machine Learning Specialization: Supervised Learning, Advanced Learning Algorithms, Unsupervised Learning · Stanford Online / DeepLearning.AI (Andrew Ng).
- Self-Directed Deep Learning Roadmap: Practical Deep Learning (fast.ai), Biomedical Transformers (Hugging Face / BioBERT), Parameter-Efficient Fine-Tuning (LoRA / QLoRA).