MSc Computer Science (Artificial Intelligence)
Machine Learning · Computer Vision · Medical AI · Data Science
LinkedIn · Research publication
I am an MSc Computer Science (Artificial Intelligence) student at the University of Nottingham with a background in machine learning, computer vision, medical-image analysis, and scalable data processing.
Before beginning my MSc, I completed a BSc in Computer Science and Engineering at BRAC University and worked as a Graduate Research Assistant at its CVIS Lab. I enjoy turning research questions into measurable systems: designing experiments, comparing models, analysing trade-offs, and communicating the results clearly.
I am currently open to graduate and early-career opportunities across AI/ML engineering, data science, computer vision, and applied research.
| Project | What I worked on | Outcome |
|---|---|---|
| 3D Brain Image Segmentation | Helped adapt and evaluate a tiled 3D convolutional neural network for memory-efficient medical-image segmentation. | Co-authored a peer-reviewed SPIE publication. |
| Crop and Weed Semantic Segmentation | Built and tuned a compact MATLAB encoder-decoder CNN using only 50 labelled image-mask pairs. | Increased weed boundary F-score from 0.661 to 0.768 and crop accuracy from 0.798 to 0.893. |
| Breast Cancer Outcome Prediction | Reconstructed and audited classical ML, TensorFlow, and PyTorch workflows for pCR classification and relapse-free-survival prediction. | SVM holdout ROC-AUC: 0.658; random-forest CV MAE: 20.60 months. |
| Distributed Fraud Detection | Implemented class-weighted Global AdaBoost around Spark ML weak learners, with distributed row-weight updates and validation-only threshold tuning. | Test fraud F1: 0.9788; PR-AUC: 0.9877 across 6.36M synthetic transactions. |
| NHS Admissions Visual Analytics | Prepared multi-year NHS admissions data with Excel and Power Query and designed an interactive hierarchical Tableau treemap. | Exposed patterns across ICD categories, emergency admissions, age bands, and time. |
| Milepost Courier | Developed a courier demo with React, Express, GraphQL and MongoDB for itemised quotations, bookings, delivery milestones and proof-of-delivery uploads. | Containerised with Docker Compose, with client and operations dashboards and automated API and desktop/mobile browser tests. |
- Computer vision and medical AI: segmentation, CNNs, 3D imaging, limited-data learning
- Machine learning: classification, regression, imbalanced learning, model evaluation
- Scalable AI: PySpark, distributed model training, runtime and partition analysis
- Applied data science: exploratory analysis, visual analytics, reproducible experimentation
- Full-stack development: React interfaces, GraphQL APIs, database-backed workflows and containerised development
| Area | Technologies |
|---|---|
| Languages | Python, JavaScript, SQL, C, C++, MATLAB |
| ML and AI | PyTorch, TensorFlow, scikit-learn |
| Data and distributed systems | PySpark, pandas, NumPy, Tableau, Power Query |
| Web applications | React, Node.js, Express, GraphQL, MongoDB |
| Containers | Docker, Docker Compose |
| Research and collaboration | Git, Jupyter, LaTeX, experiment design, technical writing |
I am developing my MSc research into a series of clear and reproducible case studies that go beyond presenting final results. My aim is to document the complete research process, including the rationale behind key design decisions, experimental methodology, alternative approaches explored, observed limitations, and the lessons learned throughout the investigation.
A central focus of this work is the analytical evaluation of privacy, security, and information disclosure risks in federated learning for medical imaging. In particular, I am interested in understanding the extent to which federated learning protects sensitive medical data in practice, the circumstances under which information may still be exposed, and the trade-offs between privacy, security, model performance, and practical usability.
By presenting the work in this way, I aim to produce case studies that are technically rigorous, transparent, and reproducible, while also providing insight into the reasoning, experimentation, and challenges that shaped the final outcomes.
If you are working on thoughtful applications of AI or hiring for graduate AI/ML roles, feel free to connect with me on LinkedIn.