Welcome to the comprehensive 5-Day Generative AI Intensive Course by Google. This course is designed to equip participants with practical skills in using Google's cutting-edge Generative AI tools and APIs to build advanced AI applications.
- Prerequisites
- Technical Requirements
- Day 1 - Prompt Engineering
- Day 2 - Embeddings and Vector Stores/Databases
- Day 3 - Agents
- Day 4 - Domain-Specific LLMs
- Day 5 - MLOps for Generative AI
- Bonus Content - Extra API Features
- Kaggle Account: Ensure your account is phone-verified.
- Google API Key: Obtain from Google AI Studio.
- Basic Python Knowledge: Familiarity with Python programming is essential.
- Internet Connectivity: Required for API access and hands-on exercises.
- Python Version: Python 3.10 or later.
- Libraries: Install the following via pip:
google-generativeai>=0.8.3
tensorflow
keras
langgraph
- Environment: Use a Kaggle Notebook or your preferred Python IDE.
Master the Gemini API and advanced prompting techniques to enhance LLM interactions.
Key Topics:
- Introduction to Gemini API setup in Kaggle.
- Prompting methods:
- Zero-shot, Few-shot, and Chain-of-Thought (CoT).
- Advanced frameworks like ReAct.
- Parameters: Temperature, top-k, top-p, and output control.
- Code-related capabilities:
- Code generation, execution, and explanation.
- Practical tools: TextFX, SQL Talk.
Outcome: Leverage prompting techniques and the Gemini API effectively for various tasks.
Deep dive into embeddings and Retrieval-Augmented Generation (RAG).
Key Topics:
- Types of embeddings: Text, image, and multimodal.
- Semantic representation and vector space analysis.
- Classification using Keras.
- Similarity scoring and optimizing vector search.
- RAG for document-based Q&A.
Outcome: Harness embeddings and RAG for building context-aware AI solutions.
Learn to build intelligent agents and implement function calling using the Gemini API.
Key Topics:
- Cognitive architectures and tool integration.
- Stateful applications with LangGraph:
- Example: BaristaBot for café orders.
- Frameworks: ReAct and other decision-making techniques.
- Practical function calling for AI-system integration.
Outcome: Design and deploy autonomous agents capable of real-world interactions.
Specialize LLMs for specific applications through search grounding and fine-tuning.
Key Topics:
- Connecting LLMs to verifiable sources:
- Static and dynamic query grounding.
- Fine-tuning models:
- Example: Classifying the 20 Newsgroups dataset.
- Performance evaluation, parameter-efficient tuning, and token management.
Outcome: Create domain-specific AI solutions tailored to specialized fields.
Operationalize your GenAI solutions with MLOps best practices.
Key Topics:
- Deployment using Vertex AI:
- Continuous evaluation, model monitoring, and governance.
- GenAI Starter Pack:
- FastAPI server, interactive UI playground, CI/CD with Terraform.
- RAG implementation patterns and observability frameworks:
- Tracking user interactions in BigQuery and visualizing with Looker Studio.
Outcome: Build production-ready GenAI applications with enterprise-grade reliability.
Explore the extended capabilities of the Gemini API.
Key Topics:
- Multimodal interactions: Text, images, audio, and video.
- File API: Handling large files (up to 2M tokens).
- Context caching for repetitive queries, reducing costs and improving efficiency.
- Streaming capabilities for real-time responses.
Outcome: Gain advanced proficiency in leveraging Gemini API for complex, multimodal AI solutions.
To begin the course, ensure all prerequisites are met and set up your environment according to the technical requirements.
Happy learning!