01. Prompting and structured output
Instruction design, role prompting, schema-constrained responses, and prompt debugging.
02. LLM deployment and production issues
Latency, throughput, observability, guardrails, caching, and cost control in real systems.
03. Fine-tuning and parameter-efficient tuning
When to fine-tune, LoRA-style adapters, data formatting, and evaluation.
04. Quantization and model optimization
Compression, inference efficiency, hardware trade-offs, and quality loss management.
05. Retrieval-Augmented Generation (RAG)
Grounding generation with external knowledge using retrieval pipelines.
06. Vector databases
Approximate nearest neighbor search, indexing, filtering, and storage design.
07. Embeddings
Dense representations, similarity, domain adaptation, and practical evaluation.
08. Document chunking and indexing
Splitting, metadata design, and index construction for retrieval quality.
09. Information retrieval
Ranking, lexical retrieval, evaluation, and hybrid search strategies.
10. Hallucination reduction
Detection, prevention, grounding, and response shaping for higher reliability.
11. Multilingual NLP
Cross-lingual transfer, multilingual embeddings, and language-specific pitfalls.
12. Text preprocessing and normalization
Cleaning, standardization, Unicode handling, and pipeline design.
13. Bias, privacy, and responsible AI
Risk management, governance, privacy, fairness, and safe deployment.
14. Core NLP concepts
Tokens, corpora, ambiguity, syntax, semantics, and task framing.
15. Classical NLP techniques
Bag-of-words, n-grams, TF-IDF, RNNs, CNNs, Seq2Seq, and Transformers.
16. Text representation methods
BoW, TF-IDF, word embeddings, sentence embeddings, and document embeddings.
17. Tokenization and subword methods
Wordpiece, BPE, unigram tokenization, and vocabulary construction trade-offs.
18. Word embeddings
Word2Vec, GloVe, fastText, analogies, and limitations of static embeddings.
19. Sequence models
Autoregressive modeling, sequence labeling, and temporal dependence.
20. RNNs and Seq2Seq
Recurrent modeling, encoder-decoder systems, and practical weaknesses.
21. Transformers
Modern language modeling architecture, scaling, and training mechanics.
22. BERT and transformer pretraining
Masked language modeling, pretraining objectives, and transfer learning.
23. Attention mechanisms
Soft attention, self-attention, cross-attention, and computational trade-offs.
24. Machine translation
Translation pipelines, alignment, decoding, and multilingual evaluation.
25. NLP evaluation metrics
BLEU, ROUGE, perplexity, WER, accuracy, precision, recall, and F1.
26. Interpretability in transformers
Attention inspection, probing, attribution, and limits of mechanistic claims.
27. Scaling laws in NLP
How performance changes with data, parameters, and compute budgets.
28. Large Language Models (LLMs)
Capabilities, limitations, training pipeline, deployment patterns, and future directions.
29. Classification metrics: accuracy, precision, recall, and F1
What each metric measures, when accuracy misleads, the precision-recall trade-off, F-beta, and multi-class averaging.
30. Activation functions and backpropagation
What activation functions are, ReLU, sigmoid, and softmax explained, and how backpropagation trains a network using the chain rule.
31. My Notes
Personal study notes: LLM fine-tuning, CNNs, RNNs, Transformers, attention, hyperparameters, KV cache, and more.
31b. Notes Exam
50 multiple-choice questions covering neural networks, attention, embeddings, training, and interview-focused topics.
32. Loss Functions
MSE, MAE, Huber, binary cross-entropy, categorical cross-entropy, hinge, KL divergence, contrastive, and triplet loss.
33. Deep Learning Fundamentals
Neurons, activation functions, backpropagation, CNNs, ResNet, transfer learning, Transformers, BERT, GPT, and generative models.