Applied NLP & Large Language Models
Multiple Institutions
Applied NLP & Large Language Models
Course Overview
This advanced course takes ML engineers and data scientists from transformer fundamentals to production-grade LLM applications. The emphasis is applied: by the end, students have built a retrieval-augmented system, fine-tuned a model on domain data, and implemented evaluation harnesses that catch hallucination and regression.
What You’ll Learn
- The transformer architecture in depth: attention, positional encoding, and scaling behavior
- Retrieval-augmented generation (RAG): chunking, embeddings, vector search, and grounding
- Fine-tuning and adaptation: full fine-tuning, LoRA/PEFT, and instruction tuning
- Evaluation: building automated harnesses for accuracy, hallucination rate, and instruction-following
- Deployment patterns: latency, cost, caching, and safety guardrails
Module Breakdown
Weeks 1–2: Foundations
Transformer internals, tokenization, and the pretraining/fine-tuning paradigm. Students implement attention from scratch.
Weeks 3–4: Retrieval-Augmented Generation
Embedding models, vector databases, retrieval strategies, and grounding techniques to reduce hallucination.
Weeks 5–6: Adaptation & Fine-tuning
Parameter-efficient fine-tuning, instruction tuning, and dataset construction for domain adaptation.
Weeks 7–8: Evaluation & Deployment
Automated evaluation harnesses, safety guardrails, and a capstone production application.
Who This Course Is For
Practicing ML engineers and data scientists comfortable with Python and PyTorch who want to move beyond API calls into building, adapting, and rigorously evaluating LLM systems.
Assessment
Four hands-on labs (RAG, fine-tuning, evaluation, deployment) and a capstone LLM application with a written evaluation report.