PRACTICAL AI FOR DEVS

Curriculum and structure

This course offers a personalized learning experience, featuring weekly lectures, office hours, and a mix of core exercises and projects to help you meet your AI goals.

• Personalized learning plan based on prior experience and goals
• Combination of core and optional exercises each week
• One live lecture per week, repeated twice and recorded
• Drop-in office hours with JV each week, at least one in the evening
• Mix of prepared curriculum, individual projects, and group projects
• Course taught in Python, no prior Python knowledge required

Pre-Course: Foundations for Success

In this preparatory phase, you'll get acquainted with various AI coding assistants and learn how to run Local Language Models (LLMs) on your own machine.

• AI coding assistants
  • GitHub Copilot, Codeium, Tabnine, Ghostwriter, CodeWhisperer, Cursor, Aider, Open Interpreter
  • Evaluation and comparisons
• AI-powered learning
• Running local LLMs

Week 1: Foundations and Interfaces

This week provides an introduction to the ethics, key players in AI infrastructure and models, and hands-on experience with APIs like OpenAI.

• Ethics, pitfalls, and possibilities
• Overview of some providers
  • Models: OpenAI, Anthropic, Cohere, Google, Huggingface
  • Infrastructure: Azure, AWS, Google Cloud, HF Spaces, Runpod
• OpenAI API
  • Temperature
  • Function calling
  • Building plugins
  • Compatible APIs
• Developing against a local LLM
• OpenRouter to easily access many models
• Google Colab and Jupyter Notebooks

Week 2: Advanced Prompting and Frameworks

This week focuses on mastering advanced prompting techniques and introduces you to academic research in the field of AI.

• Langchain framework overview
• Advanced prompting techniques
  • Zero shot, one shot, few shot
  • Step-by-step instruction
  • Chain of thought, tree of thought, graph of thought
  • Persona adoption
  • Mix of experts
• Huggingface Transformers introduction
• Strategies for engaging with academic research
• Researching and evaluating models

Week 3: Images, Voice, Video, and Translation

This week is dedicated to exploring the wide range of AI applications in multimedia, from generating images and transcribing voice to creating avatars.

• Text to image generation
• Transcription with Whisper
• Text to speech
• Voice cloning
• Avatar generation

Week 4: Retrieval Augmented Generation

This week lets you dive into the mechanics of knowledge retrieval, utilizing vector databases and augmented generation, while learning how to monitor and analyze these processes.

• Retrieval augmented generation
  • Vector databases
  • Embedding functions
  • Open vs closed providers
  • Limitations
• Knowledge graphs
• Observability and analytics (Phoenix and Langfuse)
• Exploring GptCache for caching

Week 5: Fine-Tuning and Deployment

In this week, you'll delve into the intricacies of fine-tuning models, managing training data, and deploying AI models effectively.

• Fine-tuning models
• Open (e.g., Llama 2) vs closed (OpenAI)
• Managing training data
• Synthetic training data
• Evaluation and metrics
• Deploying models

Week 6: Agent-Based AI Systems

This week guides you through the design and functioning of AI-powered agents, offering a comparative study of different frameworks and discussions on safety and alignment.

• Overview of some popular implementations
  • Auto-GPT, AgentGPT, SuperAGI, Botpress, MetaGPT, DevOpsGPT, GPT-Researcher
• Comparison of agent frameworks
  • Langchain, Haystack, Llama index
• Tools and monitoring
• Human supervision
• Recursive self-improvement, replication, and autonomous sourcing of compute **

** (Not really; we will focus extensively on alignment and safety)

Week 7: Culminating Projects

The course wraps up with this final week, where you'll apply all the knowledge and skills you've gained in individual or group projects.

Note: AI is moving fast, so our curriculum does too. Expect changes as we go.