Pathway 8: "I'm an NLP Engineer Transitioning to LLMs" (NLP / Text Mining Professional)

Target audience: NLP professionals who know spaCy, NLTK, BERT fine-tuning, NER, and text classification

Goal: Understand how LLMs change the NLP landscape, when to use prompting vs. fine-tuning vs. classical NLP, and how to migrate existing pipelines to LLM-augmented architectures.

Chapter Guide

• Skip Ch 00: ML and PyTorch Foundations (you know this) you already know this material
• Skip Ch 01: NLP and Text Representation (you know this) you already know this material
• Skip Ch 02: Tokenization (you know this) you already know this material
• Skim Ch 03: Attention (review if needed) quick review of attention for LLM context
• Focus Ch 04: The Transformer Architecture (deeper than BERT) deeper than BERT: full decoder-only architecture
• Focus Ch 05: Decoding Strategies autoregressive generation replaces your pipelines
• Focus Ch 06: Pre-training and Scaling Laws understand why scale changes everything
• Skim Ch 07: The Modern LLM Landscape survey of models replacing your NER and classification
• Focus Ch 10: Working with LLM APIs the new interface: APIs replace custom models
• Focus Ch 11: Prompt Engineering prompting replaces much of your feature engineering
• Focus Ch 12: Hybrid ML+LLM Architectures keep your best NLP models, augment with LLMs
• Focus Ch 14: Fine-Tuning Fundamentals when prompting is not enough, fine-tune
• Focus Ch 15: PEFT (LoRA, QLoRA) efficient adaptation for domain-specific tasks
• Skim Ch 17: Alignment (RLHF, DPO) understand how models learn to follow instructions
• Focus Ch 19: Embeddings and Vector Databases modern embeddings replace your TF-IDF and BERT vectors
• Focus Ch 20: RAG RAG replaces extractive QA and search pipelines
• Skim Ch 21: Conversational AI dialogue systems now powered by LLMs
• Focus Ch 29: Evaluation and Experiment Design evaluate LLM replacements against your NLP baselines
• Skim Ch 34: Emerging Architectures state-space models and alternatives to transformers
• Optional Ch 35: AI and Society broader context on AI impact and open problems