Machine learning & data science
Useful analysis starts with a question that can be investigated using the available data. My work and technical interests in data science and machine learning include examining datasets, preparing inputs and understanding what a model's results can and cannot show.
How I approach it
I use Python and SQL to inspect, clean and organize data before choosing an analysis or modeling approach. For machine learning experiments, PyTorch supports model development, while comparison and evaluation help reveal where performance changes across the data. I pay attention to missing values, assumptions and the limits of the sample so conclusions remain tied to the evidence. When a chart or written explanation is the clearest result, the work does not need a model.
The output depends on the question: it could be an exploratory analysis, a prepared dataset, a model experiment or an explanation of findings and limitations. This discipline can also connect with software and automation work, such as making analysis repeatable or exposing a useful result through an interface. The method is chosen to suit the data and decision at hand.
What I can help with
- Exploratory analysis and data visualization
- Data cleaning, preparation and SQL queries
- Supervised or unsupervised model experiments
- Model comparison, evaluation and result explanation
Tools & methods
Python · PyTorch · SQL · SQLite · MATLAB
Frequently asked questions
Do I need a huge dataset to start?
No. Useful analysis starts with a question, not a data size. A small, well-understood dataset often answers a well-posed question better than a large messy one.
Can you explain the results to a non-technical team?
Explaining findings and their limitations is part of the deliverable. A result that cannot be understood cannot be acted on.
What tools do you use?
Python and SQL for data preparation, PyTorch for model experiments, and standard visualisation tools for communicating results.