two-tower

Implementation of a Hybrid Two-Tower Architecture for Music Recommendation. Solves cold-start problems by integrating content-based multimodal encoders with sequential user modeling.

Recipe recommender using Two-Tower retrieval, LightGBM ranking, and MMR diversification

Production-grade ML recommendation engine: Kafka → Flink → Two-Tower PyTorch + FAISS ANN + LinUCB Bandit → sub-100ms FastAPI serving with live React dashboard

Production-ready personalized recommender: two-tower retrieval, CatBoost ranking, optional LLM rerank. FastAPI + Qdrant + MLflow. Workshop by learnwithparam.com

Production-grade MovieLens-25M recommender: two-stage retrieval + re-ranking with a full MLOps lifecycle (Spark features, MLflow, FAISS, FastAPI on Kubernetes, A/B testing, drift detection, Prometheus/Grafana).

A distributed ML recommendation system — real-time streaming, multi-node distributed training, and fault-tolerant, scalable serving.

Short-video recommendation reproduction: Two-Tower recall, Faiss TopK, time split, negative sampling, Recall@50/NDCG@50, ablation and badcases.

🤖 Explore and optimize rewards with Bandexa, a PyTorch-native library for Neural-Linear Thompson Sampling in contextual bandits.

LLM integration research for grocery recommendation — two-tower retrieval (PyTorch+FAISS) + LightGBM reranking + 5 Claude integration experiments

Multi-tenant two-stage movie recommender (item-item / two-tower candidates + LightGBM ranker) on MovieLens 25M — FastAPI serving behind Keycloak auth and Postgres RLS, Feast features, MLflow, k6-gated p99 < 100 ms SLO, ADR-documented.

Production-oriented personalized news retrieval and ranking on Microsoft MIND

Hybrid recommender for MovieLens-25M combining collaborative filtering (matrix factorization) with item-side content features (genres) via a two-tower architecture, ANN-based candidate retrieval with FAISS, and a lightweight MLP re-ranker on top.

Movie recommender: TensorFlow two-tower retrieval + ranker, FAISS serving, offline A/B over model variants.

Two-tower reading recommender plus an off-policy 'opportunity lab' — retrieval, LambdaMART reranking and causal evaluation.

Two-tower retrieval recommender on MovieLens with FAISS HNSW serving and diversity reranking.

Two-stage retrieval→ ranking recommender system. Two-Tower + MMoE-DeepFM, zero train/serve skew, ONNX serving. Runnable end-to-end.

PyTorch two-tower (dual-encoder) retrieval recommender with in-batch negatives and Recall@K/NDCG evaluation.

Real-time AI recommendation platform (Netflix-grade): two-stage retrieval, Kafka feature pipelines, FAISS vector search, and full MLOps. Cinematic Next.js site + Python platform source.

Two-stage short-video recommender (TikTok/Monolith-style): two-tower retrieval + a LightGBM ranker over real KuaiRec logs (12.5M interactions). Leakage-safe temporal eval; retrieval beats popularity at every k; ranking AUC 0.83; ~1ms /recommend.

Which repos would you have starred already, if you had seen them? Two-tower retrieval over public star histories, LLM fingerprints for cold start, a librarian you can talk to. Built on Hopsworks. #012 awesome-ml-systems