Chris founded bitcrowd in 2010 as a rails company, since 2016, bitcrowd mainly works with elixir projects. Their open source elixir projects include https://github.com/bitcrowd/carbonite and https://github.com/bitcrowd/chromic_pdf.
When ChatGPT had its first coding interview with Chris, he was fascinated to encounter a new form of thinking and coding, reminiscent of his fascination with BASIC on his first computer at the age of 12.
Since then, he has been researching the opportunities, dangers, and influence of AI. bitcrowd, his Elixir consultancy, has built various ML projects since 2021.
Aside from the hype around code generation, the potential of information retrieval, especially retrieval augmented generation (RAG), is often overlooked. Here, LLMs are becoming the industry standard for retrieving information from documents. Embedding Models allow the retrieval of information based on imprecise search terms in a way that was not possible with full text indices.
However, the application of RAG techniques on code bases is much less common. This talk sheds some light on the current state of retrieval augmented generation, how easy it is to use, and how it can empower development teams.
We will have a look at the general techniques, what online models are capable of, and what performance can be expected from local setups. We will explore the state of RAG support for Elixir, both as input-codebase, and as processing entity. We will further see what influence the coding style and the ingestion process has on the usefulness of the results.
Key Takeaways:
Target Audience: