Build a Hybrid Recommender for a Niche Consumer-AI Music App
Overview
What this challenge is about.
You receive listening events (around 240 million plays) plus a content embedding per track (audio + curator tags). Build a collaborative filtering model (ALS or implicit-feedback matrix factorization) and a content-based fallback for cold-start tracks, then combine them in a hybrid scorer. Evaluate offline using temporal split (train on first 16 months, test on last 2) with NDCG@10 and catalog coverage as metrics. Propose an A/B test design including primary metric, sample-size calculation, and guardrail metrics.
The Brief
What you'll do, and what you'll demonstrate.
Beat the current heuristic next-song queue by 15% on offline NDCG@10 with a hybrid collaborative + content model, and design the online A/B test.
Earning criteria — what you'll demonstrate
- Implement and tune collaborative filtering on implicit feedback
- Combine collaborative and content signals into a hybrid scorer
- Evaluate recommenders with temporal splits and ranking metrics
- Design a sound A/B test for a recommendation surface
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Machine Learning Engineer
Hybrid recommenders shipped behind A/B tests are the most common production-ML pattern at consumer companies; this challenge mirrors the exact workflow.
This challenge sharpens
- recommender-systems
- collaborative-filtering
- python
Data Scientist
Designing the A/B test with sample-size and guardrails is core junior-data-scientist work on a growth or experimentation team.
This challenge sharpens
- ab-testing
- evaluation
- recommender-systems
Applied AI Scientist
Owning the hybrid scoring weights and the cold-start fallback is bread and butter applied AI work in consumer products.
This challenge sharpens
- content-based-filtering
- collaborative-filtering
- evaluation