According to @DeepLearningAI, a new short course with Qdrant teaches multi-vector image retrieval that directly matches text tokens to image patches and outperforms single-vector methods (source: DeepLearning.AI on X, Dec 10, 2025). According to @DeepLearningAI, the curriculum includes implementing ColBERT to understand multi-vector search fundamentals (source: DeepLearning.AI on X, Dec 10, 2025). According to @DeepLearningAI, learners will apply ColPali for patch-level image retrieval and reduce memory footprint via quantization and pooling (source: DeepLearning.AI on X, Dec 10, 2025). According to @DeepLearningAI, the course uses MUVERA to enable fast HNSW search and concludes with a full multimodal RAG pipeline built on ColPali and MUVERA (source: DeepLearning.AI on X, Dec 10, 2025).