- Miłosz Mizak
- Marcin Jarczewski
This project is a part of the Machine Learning Engineering at Warsaw University of Technology course aimed at developing a practical solution for "Pozytywka," an online music streaming service. As analysts, we step into the role of tackling a vaguely described task, requiring us to specify details for implementation. The challenge involves understanding the problem, analyzing data, and sometimes negotiating with management (tutor) to ensure the models are production-ready and future-proof for subsequent versions.
"Pozytywka" collects data crucial for this project, including:
- A list of available artists and music tracks
- A user database
- User session history
- Technical information regarding the caching level of individual tracks
Extend the "Pozytywka" service by generating popular playlists – sets of matching songs tailored to capture the interest of a broad audience. This initiative aims to enhance user engagement by offering compilations based on the most popular music genres, updated weekly with 10 to 20 songs each.
- Define the business problem, modeling tasks, assumptions, and success criteria.
- Analyze the provided data to assess sufficiency for task realization, identifying any gaps or requirements for additional data.
Report available here (in Polish).
- Model Development:
- Develop a baseline model (the simplest possible for the given task).
- Develop an advanced target model.
- Report detailing the model building process and comparing results.
- Application Implementation:
- Implement an application (as a microservice) that:
- Serves predictions using the developed model.
- Conducts an A/B experiment comparing both models and collects data for later quality assessment.
- Implement an application (as a microservice) that:
- Demonstration Materials:
- Provide materials showing the implementation is functional.
Report available here (in Polish).
Project based on the cookiecutter data science project template. #cookiecutterdatascience