Building Intelligent Recommendation Systems (NBIRS)

Deep learning-based recommender systems are the secret ingredient behind personalized online experiences and powerful decision-support tools in retail, entertainment, healthcare, finance, and other industries.

Recommender systems work by understanding the preferences, previous decisions, and other characteristics of many people. For example, recommenders can predict the types of movies an individual will enjoy based on the movies they’ve previously watched and the languages they understand. Training a neural network to generalize this mountain of data and quickly provide specific recommendations for similar individuals or situations requires massive amounts of computation, which can be accelerated dramatically by GPUs. Organizations seeking to provide more delightful user experiences, deeper engagement with their customers, and better informed decisions can realize tremendous value by applying properly designed and trained recommender systems.

This workshop covers the fundamental tools and techniques for building highly effective recommender systems, as well as how to deploy GPU-accelerated solutions for real-time recommendations.

Learning Objectives

By participating in this workshop, you’ll learn how to:

  • Build a content-based recommender system using the open-source cuDF library and Apache Arrow
  • Construct a collaborative filtering recommender system using alternating least squares (ALS) and CuPy
  • Design a wide and deep neural network using TensorFlow 2 and NVTabular data loaders to create a hybrid recommender system
  • Optimize performance for both training and inference using large, sparse datasets
  • Deploy a recommender model as a high-performance web service


  • Intermediate knowledge of Python, including understanding of list comprehension
  • Data science experience using Python
  • Familiarity with NumPy and matrix mathematics

Suggested resources to satisfy prerequisites: Python Tutorial, Kaggle Learn Machine Learning Course, NumPy Tutorial


CuDF, CuPy, TensorFlow 2, NVIDIA Merlin, NVTabular, and NVIDIA Triton™ Inference Server

Assessment Type:

Skills-based coding assessments evaluate students’ ability to debug and correct production-quality recommendation pipelines.


Upon successful completion of the assessment, participants will receive an NVIDIA DLI certificate to recognize their subject matter competency and support professional career growth.

Hardware Requirements:

Desktop or laptop computer capable of running the latest version of Chrome or Firefox. Each participant will be provided with dedicated access to a fully configured, GPU-accelerated server in the cloud.

Mostra dettagli

Workshop Outline


  • Meet the instructor.
  • Create an account at

Matrix-Based Recommender Systems

Implement collaborative filtering with singular value decomposition (SVD):

  • Read sparse data into a GPU using CuPy.
  • Perform ALS efficiently with NumPy broadcasting rules.
  • Build a content-based filter with cuDF.

Training Wide and Deep Recommenders

Build a wide and deep network using TensorFlow 2:

  • Build a deep network using Keras.
  • Build a wide and deep network using TensorFlow feature columns.
  • Efficiently ingest training data with NVTabular data loaders.

Challenges of Deploying Recommendation Systems to Production

Deploy a recommender system in a production environment:

  • Acquire a trained model configuration for deployment.
  • Build a container for deployment.
  • Deploy the trained model using NVIDIA Triton Inference Server.

Final Review

  • Review key learnings and answer questions.
  • Learn to build your own training environment from the DLI base environment container.
  • Complete the assessment and earn a certificate.
  • Take the workshop survey.