Good to see you, I'm Smyrna.

In my experience as a translator, then as a business analyst, I kept running into the same situation: useful material scattered across folders with cryptic filenames, unicode issues, weird metadata, a mix of image formats and old slide decks — hours spent manually assembling reference for a task that in itself takes fifteen minutes. In trying to manage large volumes of information with no effective way to navigate it, I got pulled toward NLP and RAG naturally and learned the full process for turning unstructured data into queryable document databases.

I'm currently applying all of this to Anima Mundi, my Obsidian PKM vault — ingestion, cleaning, vector embeddings, evaluating different models for semantic accuracy and discovery of connections. More recently I've been following the Recursive Language Model paradigm and thinking about how it maps onto the retrieval problem. The personal motivation is simple enough: I accumulate information faster than I can internalise or process it, so building infrastructure that supports continuous recall and adapts to my particular way of thinking and mind mapping was an obvious turn to take.