AI Foundations

Understand the major foundation models, their capabilities, and how to interact with them via chat and API.

• What is a Large Language Model (LLM): training data, parameters, tokenization
• ChatGPT and GPT-4o: OpenAI's flagship model, capabilities, limitations
• Claude (Anthropic): model family (Haiku, Sonnet, Opus), extended thinking
• Gemini (Google): model variants, multimodal capabilities, Gemini vs PaLM
• Open-weight models: Llama 3, Mistral, Phi, Falcon — when to use vs hosted
• Model context windows: tokens, context limits, long-context models
• Chat vs completion API: system/user/assistant roles, message history
• Multimodal models: vision inputs, image generation, audio (Whisper, TTS)
• Model benchmarks: MMLU, HumanEval, HELM — what they measure
• Choosing a model: cost, latency, capability, context window tradeoffs

Free Foundation Models & Interfaces practice questions →Foundation Models & Interfaces study guide →

Design effective prompts for LLMs using established techniques for accuracy and reliability.

• Zero-shot prompting: direct instruction without examples
• Few-shot prompting: providing examples to guide output format
• Chain-of-thought (CoT): ask the model to reason step by step
• System prompts: setting persona, constraints, output format
• Temperature and top-p: controlling randomness vs determinism
• Structured output: JSON mode, response_format, tool use for structured data
• Prompt injection: what it is, risks, mitigation strategies
• Role prompting: 'You are an expert in...' patterns
• Self-consistency: generating multiple responses, majority vote
• Iterative prompt refinement: evaluating output and improving prompts

Free Prompt Engineering practice questions →Prompt Engineering study guide →

Build retrieval-augmented generation pipelines to ground LLMs in custom data.

• RAG overview: why LLMs hallucinate, how retrieval grounds responses
• Embeddings: what they are, how similarity search works, cosine similarity
• Chunking strategies: fixed-size, sentence, semantic, recursive character
• Embedding models: text-embedding-3-small, ada-002, sentence-transformers
• Vector databases: Pinecone, Weaviate, Chroma, pgvector, Qdrant
• Retrieval methods: semantic search, keyword (BM25), hybrid search
• Augmented generation: stuffing retrieved context into the prompt
• Reranking: cross-encoder reranking, Cohere Rerank, improving precision
• Evaluation: faithfulness, answer relevance, context recall (RAGAS)
• Advanced RAG: HyDE, multi-query, parent-child chunking

Free RAG & Vector Search practice questions →RAG & Vector Search study guide →

Work with AI APIs professionally — authentication, cost management, streaming, and error handling.

• OpenAI API: authentication (API key), models endpoint, chat completions
• Anthropic API: messages endpoint, model IDs, max_tokens requirement
• Google Generative AI SDK: Gemini API, vertexai vs generativeai package
• Streaming responses: server-sent events, chunked responses, UI integration
• Rate limits: RPM, TPM, handling 429 errors, exponential backoff
• Cost management: pricing per 1M tokens (input vs output), cost estimation
• Prompt caching: Anthropic and OpenAI cache mechanisms, when to use
• Function calling / tool use: defining tools, parsing tool calls
• Batch API: async processing for large jobs, cost savings
• Model versioning: pinning models, deprecation timelines, testing upgrades

Free AI Engineering Basics practice questions →AI Engineering Basics study guide →

Use AI coding assistants effectively in your development workflow.

• GitHub Copilot: in-IDE completion, Copilot Chat, CLI, accepting/rejecting suggestions
• Cursor: AI-first editor, Cmd+K inline edit, Composer multi-file changes
• VS Code AI extensions: Copilot, Codeium, Tabnine — key differences
• Claude Code (CLI): interactive coding sessions, file editing, bash execution
• OpenAI Codex: code generation model, Codex CLI tool, strengths/limits
• AI for testing: generating unit tests, edge cases, test coverage improvement
• AI for code review: explaining PRs, finding bugs, security review
• AI for documentation: generating docstrings, README, API docs
• Context window strategies: what to include for best completions
• Prompt injection risks in coding tools: user input in generated code

Free AI-Powered Dev Tools practice questions →AI-Powered Dev Tools study guide →