E-commerce marketplaces, streaming media apps, job boards, and content feeds looking to increase user engagement and average order value (AOV) frequently search to hire a freelance recommender systems & real-time personalization programmer. I am Aldian Fazrihady, a Senior Freelance AI Application Engineer and Data Engineer available via Upwork.
Core Recommendation Engine & Personalization Capabilities
- Two-Stage Retrieval & Ranking Architecture: Retrieving top-200 candidate items in under 15ms using Two-Tower / item-to-item embeddings in `pgvector` or Qdrant, followed by LightGBM / XGBoost / cross-encoder re-ranking on real-time session features.
- Solving the Cold-Start Problem with Multimodal LLM Embeddings: Generating rich semantic embeddings from product descriptions, images, and attributes immediately upon upload so brand-new catalog items get recommended before accumulating click history.
- Real-Time Session Intent Updates: Updating user preference vectors instantaneously in Redis as a shopper clicks, scrolls, or adds items to cart.
Frequently Asked Questions
Should we use AWS Personalize or build a custom pgvector + LightGBM recommender?
A custom `pgvector` + LightGBM/LLM hybrid recommender costs 70–90% less at scale, avoids black-box vendor lock-in, and lets you enforce custom business rules (inventory margin boosting, out-of-stock suppression, diversity constraints) directly in SQL/Python.
How can we hire you to build our recommendation engine?
Contact me on Upwork.