Mental health support chatbot built by fine-tuning LLaMA models (2-7b, 3.2-1B, 68m) using LoRA and QLoRA. Covers the full pipeline—from dialogue data preprocessing to training and evaluation (BLEU, Perplexity, VADER Empathy Score).

The course covers key NLP topics—from basic methods (tokenization, Word2Vec) to modern models (Transformers, BERT, GPT), their optimization, and applications (NER, chatbots). It explores both theory (RNN, LSTM architectures) and practice (fine-tuning, quantization), concluding with a discussion on NLP trends and future directions.