Community Interest Check: LLMs from Scratch in Pure Julia

[j_u]

@mantzaris

Project Interest

1. Developing pure Julia implementations for LLM training (not just inference)
2. Building the necessary distributed training infrastructure
3. Creating efficient Julia-native attention mechanisms and optimizers
4. Leveraging Julia’s strengths

I just came across the article titled “How OpenAI uses Apache Kafka and Flink for GenAI” and the presentation titled “Building Stream Processing Platform at OpenAI,” delivered at the Current 2025 conference in London. I thought I would post them here for reference in relation to the initial post.

EDIT: I noticed I had linked to the same article twice, so I corrected it.