Technology paths
From the fundamentals to a working application. Each path is curated rather than generated: the shortest route between two technologies is not usually the one worth learning.
-
From machine learning to the Transformer
The conceptual ladder. Each rung exists because the one below it was not enough for the problems people actually had.
Machine Learning → Neural Network → Deep Learning → Attention → Transformer
-
From the Transformer to a RAG application
The path the PRD names: how the architecture underneath becomes an application that answers from your own documents.
Transformer → LLM → RAG → Qdrant
-
From raw text to retrieval
The retrieval half on its own, which is where the quality of a RAG system is actually decided.
Tokenization → Embedding → Qdrant → Reranking
-
Adapting a model to your data
How to specialise a general model without paying to retrain it, and why that became possible.
Transformer → LLM → Fine-tuning → LoRA