A side project can be a perfect vehicle for active learning, but we know that many data scientists struggle to carve out the time to start one. Even if you have some free time, settling on an idea can be a challenge: it has to be relevant to your interests, reasonably scoped, and challenging enough (but not so difficult that it becomes a painful slog).
Well, we’re here to help!
This week, we’ve gathered a strong lineup of recent hands-on tutorials that might just unlock your creativity and inspire you to venture into a new topic. Follow one (or several) of them step by step, or treat any of these recommendations as the starting point for an entirely different learning adventure. Either way, the most important thing is to choose something. You’ll sort out the details as you go.
- Hit the ground running with a data-integration pipeline. If you can’t wait to dive into some code, Marie Truong‘s tutorial on building an ELT (extract, load, transform) workflow will get you to roll up your proverbial sleeves in no time. Among other useful components, you’ll get to play with an API and find your way around BigQuery.
- Create a custom microservice. Has all the endless buzz around AI made you feel a bit disoriented? A good way to re-find your bearings is to build an app that does something tangible with this rapidly evolving technology. Mason McGough‘s project walkthrough will show you how to build a Stable Diffusion app that removes people from images—a neat idea in itself, and one that might inspire you to explore other directions, too.
- No code, just math. Mastering a complex concept is easier when its various elements become more concrete. Case in point: Aparna Dhinakaran and Jason Lopatecki’s primer on the Jensen-Shannon Divergence, where they patiently explain the math behind it as well as its practical applications in the context of monitoring model drift.
- An end-to-end guide to building a geospatial tool. If your learning style responds well to detailed, slow-and-steady guidance, give Jacky Kaub‘s series on creating a map-based application a try. It will take you from proof of concept to minimum viable product over the course of three patient articles; by the end of part 1 you’ll already have a minimalist demo in hand.
- Deep learning from scratch, anyone? Piotr Lachert‘s debut TDS post starts with a straightforward premise: "the best way to check if you really understand how neural networks learn is to implement the whole process all by yourself." Piotr goes on to do precisely that, leading readers through the implementation of basic neural networks in Python.
- Investigate the nitty-gritty of causal inference. Nazlı Alagöz‘s tutorial uses the example of a hypothetical music-streaming service to unpack the complexity of drawing causal insights, and looks closely at the limitations of the difference-in-differences (DiD) approach when dealing with data shaped by staggered treatment timing and multiple time periods.
- Even more causality, just because. If Nazlı’s post piqued your curiosity about causal inference and you’re ready to learn more about it, you’re in luck: we’ve devoted our recent March Edition to this essential topic, and you’ll find several excellent tutorials there.
Thank you for your time and your support this week! If you enjoy the work we publish (and want to access all of it), consider becoming a Medium member.
Until the next Variable,
TDS Editors
