All projects
ai agents
A recommendation engine you can evaluate
Collaborative filtering, a baseline to beat, and an honest offline evaluation.
Why it matters
Recommenders are everywhere in product work. The distinguishing skill is not the algorithm but knowing whether yours is better than recommending the most popular item.
What you build
- A popularity baseline to beat
- Collaborative filtering
- Offline evaluation with a proper split
- An honest comparison against the baseline
If you want more
- Add content-based features for cold start
- Serve it behind an API
Build it
If your model cannot beat this, you have learned something important.
- Compare against the baseline and report the truth
Material for the whole build
Shows you evaluate models against a baseline, which is the difference between machine learning and hoping.
Get a plan built around projects like this