Ariel Gamino

Ariel Gamiño is a Lead AI and Machine Learning Engineer at Athenahealth, where he works on building recommendation systems. He has master’s degrees in Software Engineering from Harvard University and Predictive Analytics from Northwestern University. Much of the work he has done leverages artificial intelligence technologies for classification and recommendation tasks. He has worked in the Software Engineering space for more than 23 years and enjoys teaching others. He also teaches a Data Analytics and Visualization boot camp at the University of Texas at Austin's McCombs School of Business.

projects by Ariel Gamino

Recommendation System with Surprise and Fast.ai

2 weeks · 7-9 hours per week average · BEGINNER

In this series of liveProjects, you’ll build recommendation systems to help suggest products to the customers of an online store. You’ll create a product rating matrix to help understand user preferences and tastes, then utilize two different libraries—Surprise and Fast.ai—to make product recommendations. You’ll learn each library’s different approach to building recommendation systems and go hands-on with different techniques for building your models.

Fast.ai

1 week · 8-10 hours per week · BEGINNER

In this liveProject, you’ll create a recommendation engine for an online store using the Fast.ai library. You’ll utilize a dot product and a neural network to come up with the latent factors in a rating matrix, and compare and contrast them to determine which is likely to deliver the best recommendations. You’ll need to select and clean your data, pick the right methods, then create the functions that you need to recommend products based on predicted ratings.

Surprise

1 week · 6-8 hours per week · BEGINNER

In this liveProject, you’ll create a product recommendation engine for an online store using collaborative filtering techniques from the Surprise library. You’ll work with Amazon review datasets to create your data corpus, and identify which would be best for a collaborative filtering recommender. You’ll then use two different approaches—neighbourhood-based and matrix factorization—to implement different solutions to the rating matrix completion problem. You’ll learn how to select and clean the necessary data for these different approaches. When you’re finished, you’ll have built a system that can predict the rating for a product a user has not yet purchased.