Scaling CLIP based Natural Language Media Search with Bumblebee and Broadway

Raj Rajhans


As organizations increasingly adopt Elixir for various applications, it has become essential to effectively implement machine learning at scale within the Elixir ecosystem. In this talk, attendees will be shown how Elixir libraries Bumblebee and Broadway can be used to tackle machine learning tasks in production, using natural language media search as a prime example.

The talk will show why Elixir is a powerful choice for machine learning applications and how its capabilities can be leveraged to build scalable, efficient solutions. As an example, we will walk through the process of indexing bulk media using Bumblebee and Nx by utilizing Broadway for batch processing, and then running inference queries in production with Bumblebee against the data indexed before.

Attendees will gain insights on how to harness the power of Elixir libraries for ML applications, enabling them to build robust machine learning powered solutions using Bumblebee, Nx for ML and Broadway for data processing.

Key Takeaways:

  1. Harnessing the power of Elixir Libraries for applications requiring machine learning, at scale.
  2. Understanding how combination of Bumblebee, Nx and Broadway can be used to deploy production grade machine learning applications.
  3. Learning how natural language media search can be implemented at scale in Elixir.

Target Audience:

  • Software engineers who have experience with or interest in Elixir and want to explore its capabilities for machine learning applications
  • Technical leads and decision-makers seeking insights into the potential of Elixir for machine learning in their organizations
  • Anyone interested in natural language media search and its practical implementation using Elixir libraries

Bumblebee, Machine Learning, Openai CLIP