Build AI apps with OpenAI, Claude, Gemini — chatbots, RAG, agents, and more. 4 focused paths from your first API call to production-ready AI systems.
Your first AI-powered code in 5 minutes. Then master the APIs, prompting, and output control that everything else builds on.
1 / 12 tutorials available
Multi-turn conversations · System prompts and personalities · Conversation history as memory · Streaming responses
JSON mode for structured output · Pydantic validation · Function calling / tool use · Batch extraction
Zero-shot and few-shot prompting · Chain-of-thought reasoning · Prompt templates · A/B testing prompts
ReAct pattern · Self-consistency voting · Prompt chaining · DSPy optimization
OpenAI vs Anthropic vs Gemini patterns · Instructor library · Nested and complex schemas · Streaming structured outputs
Tool definitions with JSON Schema · The tool-use loop · Parallel function calling · Security and sandboxing
Anthropic Python SDK · Extended thinking · 200K context window · Tool use with Claude
Google AI Studio setup · Multimodal inputs · Safety settings · Grounding with Google Search
Provider-agnostic AIClient · Smart model routing · Fallback chains · LiteLLM integration
Ollama setup and usage · Open-source models · OpenAI-compatible API · Quantization basics
Vision APIs across providers · Image understanding and OCR · Audio transcription · Multi-image comparison
Build AI that answers questions from YOUR data — with citations, accuracy, and scale.
The RAG pattern · Text chunking and embeddings · Cosine similarity search · Source attribution
Embedding generation across providers · Cosine similarity from scratch · Batch embedding · Open-source alternatives
ChromaDB collections and queries · HNSW algorithm · Metadata filtering · FAISS comparison
Chunking strategies compared · Semantic chunking · Ingestion pipeline · Chunk quality evaluation
Query rewriting and HyDE · Cross-encoder reranking · Hybrid BM25 + semantic search · Parent-child retrieval
Knowledge graph construction · Entity and relationship extraction · Graph-based retrieval · Hybrid GraphRAG
Conversation-aware query rewriting · Sliding window memory · Topic switch detection · Persistent sessions
The agentic RAG loop · Query decomposition · Self-evaluation · Multi-source routing
Multimodal embeddings · Table extraction · Chart understanding · Mixed retrieval results
Faithfulness, relevance, correctness · LLM-as-a-judge · RAGAS metrics · Regression testing
Semantic caching · Cost monitoring · Query analytics · A/B testing RAG changes
Build AI that thinks, decides, and acts autonomously.
The agent loop: think → act → observe · Function calling for tool use · Multi-step reasoning · ReAct pattern from scratch
Tool design principles · Input validation · Composable tools · Tool registries
StateGraph and conditional edges · Human-in-the-loop checkpoints · Streaming agent state · Persistence and recovery
Role-based agent design · Tasks and crews · Sequential vs hierarchical execution · CrewAI memory systems
Agents, tools, and handoffs · Built-in guardrails · Provider-agnostic usage · Framework comparison
Supervisor, pipeline, debate, swarm architectures · Agent handoffs · Shared state management · Cost management
Working, short-term, and long-term memory · Semantic and episodic memory · Memory consolidation · The CAG pattern
MCP architecture · FastMCP server building · Resources and prompts · Transport protocols
MCP client construction · Tool discovery · Multi-server connections · The MCP ecosystem
Reflection and self-improvement · Planning before acting · Reflexion pattern · Escalation and guard patterns
Task completion rate · Tool selection accuracy · Safety testing · Red-teaming agents
Input/output guardrails · Prompt injection defense · Action-level safety · Security audit checklist
Text-to-SQL and text-to-Pandas · Safe code execution · Chart generation · Natural language reporting
The plan-code-test-debug loop · Safe sandboxed execution · Test generation · Error diagnosis and self-repair
Ship AI to production: fine-tuning, evaluation, safety, deployment, and architecture.
Fine-tuning vs RAG vs prompting · LoRA and QLoRA · SFT, RLHF, and DPO · Data requirements and cost
Hugging Face + Unsloth ecosystem · LoRA configuration · Training monitoring · Model export to GGUF
LLM-as-a-judge · DeepEval framework · Regression testing · Cost-quality optimization
Input/output guardrails · Guardrails AI framework · PII detection · Content moderation
Direct and indirect injection · Multi-layer defense · Canary tokens · Architectural isolation
Streamlit chat interface · Session state management · File upload for RAG · Streamlit Cloud deployment
Streaming SSE endpoints · Authentication and rate limiting · Async AI request handling · Docker deployment
Structured logging for LLMs · Cost and quality monitoring · Trace visualization · A/B testing model changes
Semantic caching · Prompt compression · Model routing by complexity · Cost projections
Production reference architecture · Multi-tier caching · Queue-based processing · Graceful degradation
Bias detection and mitigation · Transparency and explainability · Content policies · Regulatory awareness
SLMs vs LLMs trade-offs · The router pattern · Quantization and distillation · Edge deployment
Full applications combining multiple paths. Resume-worthy portfolio projects.
RAG + conversation memory · Safety guardrails · Streaming UX · Streamlit deployment
Multi-agent pipeline · Quality gates · Research → write → review loop · Human-in-the-loop approval
Text-to-Pandas and text-to-SQL · Auto-visualization · Insight generation · Natural language reports
Document classification · Multi-page extraction · Table parsing · Batch processing with evaluation
Multi-pass code review · Bug and vulnerability detection · Auto-fix generation · Severity classification
Personal knowledge base · Long-term memory · MCP tool integration · Multi-provider routing
FastAPI + RAG full stack · Document ingestion pipeline · Caching and monitoring · Docker deployment