Dc-4nderson/TheCulture-recommendation-engine

The Culture … Recommendation Engine (Two-Tower Neural CF)

Two-tower neural collaborative filtering model that scores user-post affinity for The Culture, a fashion-focused social media app. Trained on implicit feedback (views, likes, comments, saves, shares, follows, skips, reports).

Architecture

• User tower: user embedding + archetype embedding -> MLP
• Post tower: post embedding + archetype embedding -> MLP
• Score: dot product of tower outputs + bias

Metrics (validation, Top-10)

• Precision@10: 0.5455
• Recall@10: 0.9024
• NDCG@10: 0.8983

Usage

See the inference / Hugging Face call cells in the training notebook for a full working example.

Part of The Culture ML model suite.