03Applied AI
Machine learning& computer vision
Applied AI Development
Practical experience exploring and implementing machine-learning workflows, with a focus on deep learning and computer vision.
- Type
- Applied AI
- Domain
- Computer vision
- Framework
- PyTorch
- Focus
- Data, training, evaluation
Collect and organize
Images are gathered and sorted into labelled classes, then split into a training set and a validation set the model never trains on.
data = ImageFolder("data/train", transform=tf)
loader = DataLoader(data, batch_size=32, shuffle=True)PyTorch · Python · NumPy
The problem
Teaching software to understand images takes more than a model. It depends on well-prepared data, a sound training process, and honest evaluation to know whether a model really works.
What I built
Machine-learning work in Python and PyTorch, focused on deep learning for computer vision and especially image classification. It covers the full workflow: preparing and preprocessing datasets, building and training neural networks, and evaluating how well models perform.
- Neural networks
- Deep learning fundamentals
- Image classification
- Dataset preparation
- Data preprocessing
- Model training and evaluation
- Computer-vision workflows
- PyTorch development
Technical decisions
- 01
Data comes first
Datasets are prepared and preprocessed with NumPy and Pandas before training, because a model can only learn from what it is given.
- 02
Neural networks in PyTorch
Models are defined and trained in PyTorch, keeping each experiment explicit and easy to adjust.
- 03
Evaluation is part of the work
Models are evaluated rather than assumed to work, so results can be compared and improved.
- 04
Repeatable workflows
Computer-vision work is organized as a workflow, from raw images to a trained and evaluated model, so every step can be revisited.
The computer-vision workflow
- 01
Prepare
Organize image datasets for training.
- 02
Preprocess
Clean and transform data with NumPy and Pandas.
- 03
Train
Build and train neural networks in PyTorch.
- 04
Evaluate
Measure performance and compare results.
- 05
Predict
Apply trained models to new images.
Technologies
- Python
- PyTorch
- NumPy
- Pandas
- Deep Learning
- Computer Vision
- Neural networks
- Image classification
- Data preparation
- Training and evaluation
- PyTorch workflows
What I learned
This work gave me hands-on experience with how deep learning behaves in practice: how much depends on data preparation, how training choices shape results, and how to read evaluation honestly. It is the foundation I bring to building AI into real software.
