Skip to content

0.1. Projects

What is the default learning project?

The default project of this course involves a forecasting task using the Bike Sharing Demand dataset. The objective is to predict the number of bike rentals based on variables like date and time, weather conditions, and past rental data. A reference implementation is provided to fallback on if needed.

Forecasting is a critical skill with wide-ranging applications in academia and industry, utilizing diverse machine learning techniques. This project introduces challenges such as managing data subsets to prevent data leakage, where future information could wrongly influence past predictions. Through tackling this project, you'll gain hands-on experience in structuring MLOps projects effectively, offering a solid foundation for your learning journey.

Is it possible to select a personal project instead?

Absolutely! We encourage you to dive into a project that resonates with you personally. This could be a project you're currently working on professionally, or a passion project you're eager to develop further. Opting for your own project allows you to apply improvements directly within a familiar domain, streamlining the learning process by removing the need to acquaint yourself with a new project's nuances.

How to find project ideas?

Looking for inspiration? There are several online platforms offering data science challenges, complete with datasets and clearly defined objectives:

  • Kaggle: A hub for data scientists worldwide, Kaggle provides the tools and community support needed to pursue your data science aspirations.
  • DrivenData: Hosts competitions where data scientists can address significant societal challenges through innovative predictive modeling.
  • DataCamp: Offers real-world data science competitions, allowing participants to hone their skills, win accolades, and present their solutions.

Can you work on a Large Language Model (LLM) project?

Working on projects centered around Large Language Models (LLM) and Generative AI does hold similarities with predictive ML projects, particularly in the areas of model management and code structuring. However, LLM projects also present distinct challenges. Evaluating LLMs can be more intricate, sometimes necessitating the use of external LLMs for thorough testing. Additionally, the training and fine-tuning of LLMs typically demand specific hardware, like high-memory GPUs, and adhere to different methodologies compared to conventional ML tasks.

Therefore, we recommend starting with a predictive ML project to get acquainted with fundamental MLOps practices. These core skills will then be easier to adapt and apply to LLM projects, easing the progression to these more specialized areas.