Recommendation Systems Engineer
Recommendation Engineers build the personalization and recommendation systems that drive engagement and discovery across e-commerce, streaming, social, and content platforms. Their daily English covers presenting recommendation quality metrics, discussing cold-start strategies, writing architecture proposals for candidate generation and ranking, and communicating fairness considerations. This path covers the specialized vocabulary of recommendation systems.
Topics covered
- Collaborative filtering
- Content-based & hybrid approaches
- Deep learning for recommendations
- Evaluation & A/B testing
- Cold start strategies
- Fairness & diversity
Vocabulary spotlight
4 terms every Recommendation Systems Engineer should know in English:
A recommendation approach that predicts a user's preferences based on the behaviour of similar users — "people like you also liked..."
"User-based collaborative filtering worked well for our early user base but struggled with the cold start problem for new users."
The challenge of making recommendations for new users or new items that have no interaction history yet
"We tackled the cold start problem by collecting explicit preferences during onboarding and using content-based features until behavioural signals accumulate."
Normalized Discounted Cumulative Gain — an offline evaluation metric that measures ranking quality, giving more credit to highly relevant items appearing at the top of a recommendation list
"Our new ranking model improved NDCG@10 by 4% in offline evaluation, which historically correlates with a 1-2% lift in click-through rate."
A feedback loop in recommendation systems where items that are shown more often receive more implicit feedback, leading the model to recommend them even more — amplifying popularity over relevance
"Exposure bias caused long-tail content to be systematically underrecommended, reducing content diversity for users."
📚 Vocabulary Reference
Key terms organised by category for Recommendation Systems Engineers:
Core Approaches
Architecture
Evaluation
Challenges
Recommended exercises
Real-world scenarios you'll practise
- Writing a recommendation system architecture proposal: explaining the candidate generation, ranking, and re-ranking stages and the model choices at each stage
- Presenting recommendation quality metrics to product stakeholders: translating NDCG, coverage, and serendipity into business value language
- Explaining the cold start strategy to the product team: presenting the multi-stage approach for new users from onboarding signals to collaborative filtering
- Running a fairness review: identifying and presenting the exposure bias and filter bubble risks in the current recommendation system