Machine Learning & Deep Learning Refresher

• Review key ML and DL concepts, with a focus on supervised/unsupervised learning, CNNs, RNNs, and gradient descent.

• Project: Build and evaluate a basic image classification model.

Transformers: From Neural Networks to Attention Mechanisms

• History and evolution: From neural networks to RNNs, LSTMs, and finally Transformers.

• Detailed breakdown of self-attention and transformer architecture.

• Project: Implement a simplified transformer for a text classification task.

Exploring Transformer Architectures

• Study various transformer models, including BERT, GPT, and T5, to understand their unique architectures and use cases.

• Project: Compare and fine-tune BERT and GPT models on a text-based task to see differences in performance.

Advanced Use of Pre-Trained LLMs

• Hands-on with major pre-trained LLMs (e.g., GPT-4, LLaMA) and understanding prompt engineering.

• Experiment with prompt tuning and parameter adjustments for different tasks.

• Project: Design prompts for LLMs to perform specific business-oriented tasks (e.g., summarization, Q&A).

Components of LLMs: Tokenizers and Embedding Models

• Understanding tokenization and embedding creation, with emphasis on subword tokenization.

• Project: Build a custom tokenizer and simple embedding model.

Creating a Vector Database and Integrating Search

• Step-by-step creation of a vector database for storing and searching embeddings.

• Practical implementation with libraries such as FAISS or Pinecone.

• Project: Build a mini search engine using custom embeddings and a vector database.

Retrieval-Augmented Generation (RAG) and Its Components

• Breakdown of RAG architecture, combining LLMs with vector databases for enhanced QA capabilities.

• Project: Implement a basic RAG pipeline to answer questions from a custom dataset.

Finetuning Techniques for LLMs and Embeddings

• Methods for finetuning LLMs, embeddings, and tokenizers (including LoRA, QLoRA).

• Hands-on with distributed finetuning using multi-GPU clusters.

• Project: Finetune an embedding model on domain-specific data, analyze results.

Evaluating LLM and Embedding Models

• Explore techniques for model evaluation, including perplexity, accuracy, and human evaluation for generative tasks.

• Project: Create a framework for evaluating different LLMs based on specific metrics.

Multi-GPU and Distributed Computing for Scaling Models

• Learn to distribute workloads with PyTorch DDP, Data Parallel, and FSDP.

• Project: Scale the fine-tuning of a transformer model on multiple GPUs and assess improvements.

Portfolio Projects

• Curate projects from the Learning Phase to highlight generative AI skills.

• Focus on creating GitHub repositories with well-documented code, sample outputs, and explanations.

LinkedIn & GitHub Optimization

• Update LinkedIn profile with relevant skills, project links, and descriptions of hands-on generative AI work.

• Optimize GitHub profile with showcased projects, README files, and tutorials for better visibility.

Resume Tailoring for Generative AI Roles

• Emphasize key generative AI skills and projects on the resume.

• Include certifications, contributions to open-source generative AI projects, and any notable achievements in AI.

Explaining Generative AI Projects and Contributions

• Practice presenting key projects, with a focus on the business impact and technical challenges.

• Develop short, effective responses for behavioral and technical questions regarding generative AI.

Technical Discussions on Generative AI

• Prepare for technical discussions on foundational concepts, architectures, and recent trends in the field.

• Practice explaining complex concepts (e.g., attention mechanisms, tokenization) in a simplified manner.

Role-Specific Skill Emphasis

• Highlight skills relevant to various roles in generative AI, such as model evaluation, deployment, and real-time application integration.

Mock Technical Interviews

• Conduct mock interviews focusing on system design, algorithmic skills, and detailed explanations of generative AI projects.

Case Studies and Problem-Solving Exercises

• Work on problem-solving exercises and case studies related to generative AI, such as designing a custom LLM-based solution for a business case.

Advanced Coding Challenges and Whiteboarding

• Tackle coding exercises involving transformers, NLP, and embeddings.

• Practice whiteboarding sessions to articulate solutions for architecture-related questions in generative AI.

Behavioral and Soft Skills Practice

• Prepare for behavioral interviews with a focus on communication, teamwork, and real-life examples of handling complex projects.

Market Research on Compensation

• Research industry standards for generative AI roles based on region, company size, and role requirements.

Highlighting Unique Skills

• Emphasize specialization in advanced generative AI topics (e.g., distributed finetuning, embedding creation, RAG pipeline building).

Negotiation Practice

• Practice negotiation scenarios to confidently discuss benefits, perks, and salary adjustments.