Generative AI Training Syllabus

 Generative AI

• Introduction to AI

• AI vs ML vs DL vs NLP vs Generative AI

• Generative AI principles

• Role of ML in Generative AI

• Different ML techniques (Supervised, Unsupervised, Semi-supervised, Reinforcement Learning)

• Applications in various domains

• Ethical considerations


NLP and Deep Learning

• NLP essentials

• Basic NLP tasks

• Different text classification approaches

• Frequency based: TF-IDF, BoW, N-gram

• Distribution Models: CBOW, Skipgram (Traditional approaches), Word2Vec, GloVe

• Ensemble and traditional Machine learning Models: Naive Bayes, SVM, Logistic Regression, Decision Trees

• Deep learning techniques: CNNs, RNNs, LSTMs, GRU, Transformers


Generative AI models

• Autoencoders

• VAE's and applications

• GAN's and its applications

• Different types of GAN's and applications

• Language Models and Transformer models

• Different types of Language models

• Applications of Language models

• Transformers and its architecture

• BERT, RoBERTa, GPT variations

• Applications of transformer models and Hugging face


Prompt Engineering

• What is Prompt Engineering

• Different principles of Prompt Engineering

• Types of Different Prompt Engineering techniques

• Crafting effective prompts to LLMs

• Priming Prompt

• Prompt Decomposition


Generative AI Lifecycle and LLMs working procedure

• Generative AI lifecycle

• What is RLHF

• LLM pre-training and scaling

• Different fine-tuning techniques


Different Embeddings and Techniques

• Word embeddings

• Use cases of word embeddings

• Word Embeddings: Word2Vec, GloVe, FastText

• Contextual Embeddings: ELMo, BERT, GPT

• Sentence Embeddings: Doc2Vec, Infersent, Universal Sentence Encoder

• Subword Embeddings: BPE (Byte Pair Encoding), Sentence Piece

• Use case of Embeddings


Different chunk metrics

• Chunking

• Use of chunking the document

• Traditional effective chunking techniques

• Problems and limitations with traditional chunking techniques

• Overcoming limitations of Traditional chunking

• Advanced chunking techniques:

• Character splitting

• Recursive character splitting

• Document based chunking

• Semantic Chunking

• Agentic Chunking


RAG and Advanced RA with Langchain

• What is RAG

• Main components of RAG

• High-level architecture of RAG

• Building RAG using external data sources

• Advanced RAG


Langchain for LLMs

• What is Langchain

• Core concepts of Langchain

• Components of Langchain

• How to use Langchain agents


Vector Databases

• LlamaIndex

• Vector Databases

• Why prefer Vector databases over traditional databases

• Different types of Vector databases: Open-source and Close Source

• Open-source: Chroma DB, Weaviate, Faiss, Qdrant

• Close-Source Vector Databases: Pinecone, ArangoDB, Cloud-Based Solutions


Finetuning LLMs

• Supervised Finetuning

• Repurposing-Feature Extraction

• Advanced techniques in Supervised finetuning: PEFT, LoRA, QLoRA


LLM's on Cloud

• Amazon bedrock, Azure OpenAI, GCP


LLMs Evaluation

• Text based LLMs:

• Automatic Evaluation: BLUE score, ROUGE Score, METEOR, BERT Score

• Human Evaluation: Coherence, Factuality, Originality, Engagement

• Image based LLMs:

• Automatic Evaluation: Pixel-level metrics, FID (Frechet Inception Distance), IS (Inception Score), Perceptual Quality Metrics, Diversity Metrics

• Human Evaluation: Photorealism, Style, Creativity, Cohesiveness

• Audio generation LLMs:

• Automatic Evaluation: FAD (Frechet Audio Distance), IS (Inception Score), Perceptual Audio Quality Metrics - PAQM, PAQM - SNR (Signal-to-Noise Ratio), PAQM - PESQ (Perceptual Evaluation of Speech Quality)

• Human Evaluation: Perceptual Quality - PQ, PQ - Naturalness, PQ-Fidelity, PQ - Musicality, Task specific evaluation

• Video Generation LLMs:

• Automatic Evaluation: Fréchet Video Distance (FVD), Inception Score (IS), Perceptual Quality Metrics, Motion-Based Metrics - Optical Flow Error, Content-Specific Metrics

• Human Evaluation: Visual Quality, Temporal Coherence, Content Fidelit

LLMops

• Model Deployment and Management

• Scalability and Performance Optimization

• Security and Privacy

• Monitoring and Logging

• Cost Optimization

• Model Interpretability and Explainability

• Continuous Integration and Deployment (CI/CD)

• Collaboration and Workflow Management

• Regulatory Compliance

• Disaster Recovery and Failover


Different AI Tools

• ChatGPT, Gemini, Copilot, Claude, Perplexity.

Comments