Talks

Walked through my journey of optimizing an Elixir solution to the One Billion Row Challenge from 12 minutes down to 25 seconds. Covered profiling with eFlambe and Benchee, concurrency modeling on the BEAM, and practical performance tuning strategies.

Demonstrated how to implement machine learning at scale in Elixir. Covered using OpenAI's CLIP model via Bumblebee for semantic image search, Broadway for batch-processing large media collections, and HNSWLib for nearest-neighbor lookups in production.

A beginner-friendly walkthrough of Nix for hassle-free dependency management. Covered why Nix solves reproducibility and environment consistency problems, and how to get started with it in your projects.

Showed how to build text-to-image and image-to-image search using Elixir's ML ecosystem. Walked through the full pipeline from indexing images into CLIP embeddings to querying them with natural language at inference time.

Talked about using Webpack 5 Module Federation to enable micro-frontend architectures with Create React App. Covered how to dynamically share code between independently built React applications at runtime, and workarounds for things like live reload.