Academy
Learning resources, tutorials, and courses.
LLM Makerspace
AI Makerspace - Scaling TikTok's Recommendation System: 1M Writes/Second with Hopsworks
In this makerspace, we show you how to build and operate a real-time feature streaming pipeline that writes at 1m ops/sec - TikTok Scale.
Build Your Own Private PDF Search Tool: Using RAG and Fine-Tuning in one Platform
We build an AI system that allows you to search your private PDF files.
Building a Cheque Fraud Detection and Explanation AI System using a fine-tuned LLM
We build a LLM that not only detects whether a scanned image of a cheque is a fraud, but also writes an explanation for why the cheque has been marked as fraud.
Function Calling for LLMs: RAG without a Vector Database
We look into extending RAG for LLMs to include the ability to query structured data and API calls using function calling.
Tutorials
Batch & Real-Time Machine Learning Systems; A better framework with Hopsworks Feature Store
Jim explores the machine learning pipeline framework known as FTI; Feature Training and Inference and argues on how this is a better mind map for ML systems, both batch and real-time.
Deploying Managed Hopsworks Via the UI or Terraform
Mahmoud presents the different options and details on how to deploy managed Hopsworks (the enterprise version of Hopsworks) on any clouds via the Hopsworks Control Plane or Terraform.
External Feature Groups in Hopsworks; Leverage Your Existing Data
Fabio introduces the concepts and different use cases for the external feature groups in Hopsworks; data that does not resides in Hopsworks but can still be used for machine learning, even for real-time use cases.
Feature Groups in Hopsworks: Working with Dataframes
Fabio shows how to work with feature groups in the Hopsworks feature store and how Hopsworks uses dataframe to help datascientists manipulate data.
Feature Pipelines in Production with Hopsworks: Code, Deployment & Monitoring
Fabio presents the 3 main ways to keep feature pipelines in production with Hopsworks. From the management of the code and the deployment to the monitoring.
Feature Views in Hopsworks: How to Join Feature Groups
Fabio introduces feature views in Hopsworks; a way to join feature groups together in a point in time correct manner with a simple API and avoid inconsistencies in the pipeline by using the same view for training and inference.
Generating Training Data From Features in Hopsworks
Fabio reviews the different ways to generate training data for machine learning models, from the features stored in Hopsworks.
Real-Time Inference: How To Retrieve Data From Hopsworks
A guide to how to use the Hopsworks API and Feature Views to retrieve data for real-time inference.
Streaming Feature Pipelines in Hopsworks; Making Data Available in Real-time
Fabio shows how users can create streaming feature pipelines in Hopsworks using Spark Streaming and Kafka.
Versioning of Features and Data in Hopsworks
Fabio walks through the different ways to version data and features in the Hopsworks Feature Store.
Show, Don't tell
Show, don't tell - Building Data Pipelines with AI
Lex and Aleksey demonstrate how we use LLMs to automate feature engineering and streamline ML pipelines with Hopsworks, making data transformation and deployment more efficient and scalable.
Show, Don't tell - High Availability & Robustness at Hopsworks
Antonios demonstrates the high availability and robustness capabilities of the Hopsworks platform.
Show, don't tell - Hopsworks to Kubernetes
Javier demonstrates how we migrated Hopsworks to Kubernetes, reducing deployment times significantly. With our approach, we've cut deployment from 2 hours to less than 30 minutes using a Helm install.
Get Started
What is a Feature Store for Machine Learning?
Lex explains what is a feature store in a nutshell; what does it do, why is it important, and how it helps organisations build more models faster.