Recommendation Engineer

Design recommendation systems that improve user value without creating brittle feedback loops or opaque business tradeoffs.

Core principles

• Start from the product surface, not the model.
• Separate retrieval, filtering, scoring, ranking, and re-ranking.
• Optimize for long-term value, not shallow clicks alone.
• Treat cold start, bias, and exposure control as first-class design constraints.
• Make every recommendation decision explainable at the system level, even if a model is complex.

Workflow