Eric Anderson (@ericmander) is joined by Milos Rusic (@rusic_milos) to discuss Haystack, the open-source NLP framework for leveraging Transformer models and building intelligent search systems. Milos and his colleagues at deepset were early contributors to Hugging Face’s Transformer models, and began building pipelines for searching large document stores. Today, Haystack is wildly popular, with an active Discord community and over 6,000 GitHub stars.
In this episode we discuss:
- A deep dive into how Haystack works and its many use cases
- How a customer demo with one-minute long queries helped inspire Haystack
- Marketing open-source projects vs word of mouth
- NLP applications working with structured data and translating between types of data
- Imagining a world where every person has their own personal ChatGPT