From Random Forest to LangChain Knowledge Retrieval Systems: Practical Applications of Data Analysis and LLMs

Part I Practiced in code the process of loading a PDF as page-by-page Documents, chunking it, converting the chunks into vectors through embeddings, and indexing them in a vector store. Learned the structure of converting questions into vectors, performing similarity searches (including scores, vectors, and MMR), wrapping the process as a Retriever (Runnable), and connecting it to RAG. Established practical standards for application by organizing search quality checks, security and governance (including prompt injection), and common mistakes. Part II Implemented in R the random forest algorithm, which overcomes the limitations of decision trees, and learned how to accurately interpret model results. Built the ability to read and explain key metrics such as OOB error, variable importance, and confusion matrices using iris data and real-world biomarker data. Completed an end-to-end workflow applicable directly in practice, covering hyperparameter tuning, cross-validation, and ROC curve analysis.