I'm a B.Sc. Data Science and Artificial Intelligence student at Saarland University in Germany.
I'm interested in using data science, machine learning, and software development to solve practical problems and build useful technical projects.
- Machine Learning
- Data Analysis
- Artificial Intelligence
- Software Development
- Algorithms and Data Structures
Programming
- Python
Machine Learning & AI
- PyTorch
- scikit-learn
- Hugging Face Transformers
- Machine Learning
- Deep Learning
Data Science & NLP
- pandas
- NumPy
- NLTK
- Natural Language Processing
- Data Analysis
- Statistics
Computer Vision & Speech
- Computer Vision
- Vision Transformers
- Convolutional Neural Networks
- torchaudio
- Audio Classification
- Model Robustness & Distribution Shift
Software Development
- Django
- Django REST Framework
- REST APIs
- Git & GitHub
Computer Science Foundations
- Algorithms
- Data Structures
- 馃帗 Studying Data Science and Artificial Intelligence at Saarland University
- 馃捈 Looking for Werkstudent and HiWi opportunities in Saarland
- 馃殌 Building practical projects in Data Science, Machine Learning, and Software Development
Academic computer vision research project comparing ResNet-50 and ViT-S/16 under distribution shift.
- Trained and evaluated ResNet-50 and ViT-S/16 under controlled conditions
- Evaluated robustness using ImageNet-C, ImageNet-A, and ImageNet-R
- Analysed corruption families including noise, blur, weather, and digital corruptions
- Studied the effects of AugMix and training duration
- Compared absolute and relative robustness metrics
Topics: Computer Vision 路 Vision Transformers 路 ResNet 路 Robustness 路 Distribution Shift 路 Deep Learning
Academic speech-classification project for identifying spoken languages from audio recordings.
- Fine-tuned W2V-BERT for 22-class spoken language identification
- Built the training pipeline with PyTorch, torchaudio, and Hugging Face Transformers
- Applied speed perturbation, pitch shifting, and Gaussian-noise augmentation
- Analysed model behaviour using confusion matrices and t-SNE visualisations
- Best reported validation accuracy: 36.45%
Topics: Deep Learning 路 Audio Classification 路 W2V-BERT 路 PyTorch 路 Hugging Face 路 Speech Processing
German-language NLP project analysing open-ended political survey responses.
- Compared BERT, lexicon-based sentiment analysis, and a rule-based ensemble
- Used Python, pandas, Hugging Face Transformers, NLTK, and scikit-learn
- Evaluated the methods against manually annotated responses
- Combined approach achieved 58.0% accuracy and a Macro-F1 score of 0.537
- Created visualisations for sentiment results, term frequencies, and model comparison
Topics: NLP 路 BERT 路 Sentiment Analysis 路 Python 路 Machine Learning 路 Data Analysis
Portfolio documentation for an academic Django software-engineering project.
- Extended an existing Django social-network application
- Worked on reputation and profile-update functionality
- Implemented expertise-community and timeline features
- Worked with user-similarity logic and relational data
- Used Django REST Framework and backend/API functionality
Topics: Python 路 Django 路 REST APIs 路 Backend Development 路 Databases 路 Software Engineering
- LinkedIn: Abdullah Mustafa