Master Generative AI with 159 tutorials across 12 sub-paths — from your first API call to production-ready AI agents, RAG systems, fine-tuned models, and deployed applications.
Your first AI-powered code in 5 minutes. Then master every major API — OpenAI, Claude, Gemini, Ollama — plus streaming, structured output, and cost management.
18 / 18 tutorials available
Build reusable prompt systems, implement every major reasoning technique, detect hallucinations, defend against prompt injection, and evaluate prompt quality with automated pipelines.
12 / 12 tutorials available
Master LCEL chains, prompt templates, output parsers, document loaders, text splitters, tools, memory, callbacks, and LangSmith — then build real applications.
16 / 16 tutorials available
Build RAG from scratch, then with LangChain. Benchmark embedding models, compare vector databases, master chunking, hybrid search, re-ranking, and evaluate with RAGAS.
6 / 12 tutorials available
BM25 keyword search · Reciprocal Rank Fusion · Alpha parameter tuning · Hybrid vs semantic-only vs keyword-only
HyDE hypothetical document embedding · Multi-query rephrasing · Sub-question decomposition · Step-back prompting
Bi-encoder vs cross-encoder · Hugging Face cross-encoder re-ranking · Cohere Rerank and FlashRank · Two-stage retrieval pipeline
Faithfulness and answer relevancy · Context precision and recall · RAGAS evaluation pipeline · Diagnosing retrieval vs generation issues
Anthropic contextual retrieval approach · LLM-generated context headers · Hierarchical headers as free alternative · Retrieval precision improvement
Query classification and routing · Department-level metadata filtering · Cross-department result combination · Date range and document type filtering
Corrective RAG, Self-RAG, GraphRAG with knowledge graphs, Agentic RAG, multimodal RAG, and production-ready enterprise RAG platforms.
Document relevance grading · Web search fallback · Query reformulation · CRAG pipeline with LangGraph
Adaptive retrieval decision · Self-critique for faithfulness · Reflection and regeneration loop · Self-RAG with LangGraph
Query type classification · Strategy routing (BM25, semantic, iterative) · Adaptive RAG router with LangGraph · Fixed vs adaptive strategy benchmarking
Entity and relationship extraction · Knowledge graph in Neo4j · Community detection · Hybrid graph + vector retrieval
Unstructured.io document parsing · ColPali visual embeddings · Multimodal vector store · Text-only vs multimodal RAG comparison
Dynamic tool selection for retrieval · Multi-source parallel retrieval · Iterative refinement loop · Agentic RAG with LangGraph
RAG vs context stuffing benchmark · Lost-in-the-middle at scale · Cost and latency comparison · Decision framework
Automated nightly quality checks · Time-series metric tracking · Drift detection · Quality alert system
Embedding and semantic response caching · Cost-aware retrieval · Graceful degradation · Production RAG wrapper
JWT authentication and tenant isolation · Per-tenant document upload and indexing · RAGAS evaluation per tenant · Docker Compose deployment
State management, conditional routing, checkpointing, human-in-the-loop, streaming, and multi-agent orchestration — all as a visual graph of nodes and edges.
StateGraph, nodes, and edges · Conditional routing · Compile and invoke pattern · Graph visualization
TypedDict vs Pydantic state · Annotated reducers for list accumulation · Nested state structures · State design patterns
ToolNode and tools_condition · Agent decision loop · Error recovery and retry logic · Data analysis agent
Tavily search integration · Search result analysis · Citation synthesis · Multi-step research
MemorySaver, SqliteSaver, PostgresSaver · Thread IDs for multiple conversations · Replay and inspect state · Resume across restarts
InMemoryStore and BaseStore · User preferences across conversations · Memory retrieval and injection · Memory management and deduplication
interrupt_before for approval gates · State editing before execution · Time-travel debugging · Email agent with human approval
astream_events interface · Token and tool call streaming · Agent state streaming · Streamlit chat UI integration
Natural language to SQL generation · Safe query execution and validation · matplotlib chart generation · Error recovery for failed SQL
Supervisor routing architecture · Researcher, writer, and coder specialists · Cross-agent coordination · Feedback loop for revision
Horizontal handoff between peers · Context-preserving state transfer · Dynamic routing decisions · Circular handoff safeguards
Manager → team leads → workers hierarchy · Strategic delegation · Upward reporting patterns · Hierarchical vs flat comparison
Sub-graph packaging · State mapping between parent and child · Composing sub-graphs · Reuse across applications
Send API for dynamic parallel branching · Map step for parallel processing · Reduce step for aggregation · Error handling in parallel branches
Researcher, writer, reviewer agents · Quality scoring and revision loop · Iterative refinement until quality passes · Human review before publishing
Same agent in raw Python vs LangGraph · Code complexity comparison · Debugging and testing comparison · Framework tipping point
Build with CrewAI for role-playing agents, OpenAI Agents SDK for lightweight agents, and AutoGen for multi-agent conversations. Compare all four frameworks head-to-head.
Agent roles, goals, and backstories · Task definitions and expected outputs · Sequential vs hierarchical process · Research crew: researcher + writer + editor
Custom tools for agents · Short-term, long-term, and entity memory · Knowledge base integration · Support crew with ticket history
@start, @listen, @router decorators · Conditional routing and loops · Parallel crew execution · State management between steps
Pydantic output models · Execution callbacks and logging · Claude, Gemini, and Ollama integration · FastAPI deployment
Agent class, Runner, and tool system · Handoffs between agents · Input and output guardrails · Built-in tracing
Pydantic structured agent responses · Real-time streaming events · Multi-agent topologies · MCP server integration
ConversableAgent building block · Two-agent Coder + Critic chat · Group chat with multiple agents · Termination conditions
Custom agent classes · Docker-based code execution · Human proxy agent · Autonomous coding team
Same system in 4 frameworks · Code complexity and readability comparison · Feature comparison matrix · Decision framework for framework selection
Agent unit and integration testing · Observability and failure modes · Circuit breakers and graceful degradation · Docker Compose deployment
Build MCP servers from scratch — SQLite, REST APIs, file systems, web scraping — and integrate them with Claude Desktop, LangChain, CrewAI, and OpenAI Agents SDK.
N x M integration problem · MCP client-server architecture · Resources, tools, and prompts primitives · stdio vs SSE transport
FastMCP and decorator pattern · Resources: list tables, show schema · Tools: run queries with validation · MCP Inspector and Claude Desktop testing
API endpoint to tool mapping · Authentication and header management · Response transformation for LLMs · Rate limiting and error handling
File listing, reading, and searching · Security boundaries and allowed paths · File type restrictions and size limits · Sandboxing and audit logging
Fetch and parse with BeautifulSoup · Structured data extraction · Response caching · URL allowlists and rate limiting
MCP-to-LangChain tool bridge · Multi-server agent connection · Dynamic tool discovery at runtime · Multi-source research agent
MCP servers as CrewAI tools · Native MCP integration with OpenAI SDK · Same server across 3 frameworks · Cross-framework architecture patterns
Multi-domain tool organization · Cross-module operations · Workspace MCP server · Usage monitoring and analytics
SSE and Streamable HTTP transport · JWT authentication for remote access · Docker containerization · Multi-client team access
MCP gateway with routing · Per-user, per-server permissions · Monitoring dashboard · Auto-generated tool documentation
Building an MCP client in Python · Tool discovery from MCP servers · Multi-server simultaneous connection · Error handling and reconnection
Prepare datasets, fine-tune with LoRA and QLoRA using Unsloth, align with DPO, quantize to GGUF/GPTQ/AWQ, and serve models with vLLM, TGI, and Ollama.
Prompting vs RAG vs fine-tuning spectrum · Cost/quality trade-off analysis · Hardware requirements · Decision framework for 10 scenarios
Instruction dataset formatting · Synthetic data generation with GPT-4 · Data cleaning and deduplication · Quality validation
Low-rank decomposition of weight updates · LoRA from scratch in PyTorch · Hugging Face PEFT LoRA · Rank, alpha, and target module hyperparameters
4-bit NormalFloat quantization · Double quantization · BitsAndBytes configuration · Fine-tuning on free Colab T4
Custom CUDA kernel optimizations · Unsloth vs standard training benchmark · Multi-format model export · GGUF and vLLM export
SFTTrainer configuration · Learning rate scheduling and WandB logging · Checkpoint management · LoRA adapter merging and Hub publishing
RLHF vs DPO comparison · Preference pair dataset creation · DPOTrainer with TRL · Alignment evaluation with LLM-as-Judge
FP16 to INT4 quantization · GGUF for CPU/Ollama inference · GPTQ and AWQ for GPU inference · Quality, speed, and memory benchmarking
SLERP, TIES, DARE merge strategies · MergeKit tool · Code + reasoning model merging · Merge ratio optimization
PagedAttention and continuous batching · OpenAI-compatible API server · Performance tuning · vLLM vs transformers benchmark
TGI Docker deployment · Ollama for local serving · vLLM vs TGI vs Ollama benchmark · Serving framework decision matrix
MMLU and HumanEval benchmarks · LLM-as-Judge scoring · Human evaluation protocol · Statistical significance testing
Code generation fine-tuning · Text-to-SQL fine-tuning · Medical Q&A fine-tuning · Task-specific evaluation metrics
End-to-end pipeline: data to production · QLoRA + Unsloth training · Quantize and deploy with vLLM and Ollama · Production monitoring and runbook
Phi-3, Gemma 2B, Llama 3.1 8B comparison · SLM router pattern · Edge deployment (Raspberry Pi, mobile, WASM) · Cost break-even analysis
Streaming APIs, Docker, caching, guardrails, observability, AWS/GCP deployment, autoscaling, security, compliance, and cost optimization.
1 / 16 tutorials available
FastAPI project structure · SSE and structured JSON endpoints · Pydantic request/response models · Health check and OpenAPI documentation
Multi-stage Dockerfile builds · Image size optimization · NVIDIA container toolkit for GPU · Docker Compose multi-service deployment
Redis exact-match caching · Semantic similarity caching · Response precomputation · Cache invalidation strategies
Input guardrails for injection prevention · PII detection and redaction · Content classification · NeMo Guardrails and Guardrails AI
Structured logging for LLM calls · Distributed tracing with OpenTelemetry · Latency, cost, and error rate metrics · Monitoring dashboard and alerts
Lambda, ECS, App Runner comparison · AWS Bedrock for managed LLM access · App Runner deployment · Monthly cost estimates
Cloud Run container deployment · Vertex AI model garden · Cloud Functions for serverless · GCP vs AWS cost comparison
Request queuing and rate limiting · Horizontal and vertical scaling · LLM budget limits · Load testing and verification
JWT authentication · Per-user rate limiting with Redis · Input sanitization · Audit logging for compliance
Model tiering (mini for 70% of requests) · Prompt compression · 3-layer caching · Real-time cost monitoring with kill switches
Mocking LLMs for unit tests · End-to-end integration tests · Regression test suites · CI/CD pipeline with AI-specific tests
GDPR for LLM applications · Data retention and right to deletion · AI transparency requirements · Compliance checklist
GGUF on mobile with llama.cpp · ONNX Runtime cross-platform deployment · Model selection for edge devices · Edge vs cloud decision framework
50-item production readiness checklist · API, performance, and reliability checks · Security and compliance verification · Incident response runbook
Streamlit chat interface · Session state for conversation persistence · File upload for documents · Streamlit Cloud deployment
Thirteen portfolio-grade agent projects combining RAG, tool use, multi-agent systems, streaming, and deployment — the projects that get you hired as a GenAI engineer.
Planner-researcher-analyst-writer-reviewer agents · Multi-source research (web, ArXiv, database) · Source quality scoring · Revision loop until quality passes
Docker sandbox with resource limits · Code generation from natural language · Error reading, debugging, and retry · Data visualization generation
Ticket classification and routing · RAG over product documentation · Specialist agents (billing, technical) · Escalation detection and resolution tracking
Calendar, email, notes, and task tools via MCP · Cross-session memory · Human approval gates · Intent classification and routing
6-agent pipeline (research to SEO) · Section-level quality gates · SEO optimization · Multi-format output (HTML, Markdown)
AST parsing and pattern checking · Bug and security issue detection · Severity classification · Structured review report generation
Market data fetching tools · Financial ratio calculations · Chart generation from data · Investment summary reports
Student level assessment · Adaptive difficulty adjustment · Practice question generation · Cross-session progress tracking
Reflection pattern · Planning and decomposition patterns · Reflexion (learn from failures) · Guard and escalation patterns
Task completion rate measurement · Tool selection accuracy · Cost and step efficiency tracking · Red-teaming and regression testing
Input, output, and action guardrails · Prompt injection defense for agents · Guardrails AI framework · Security audit checklist
Text-to-SQL and text-to-Pandas · Safe code execution sandbox · Chart generation from questions · Multi-step analysis pipeline
Plan-code-test-debug iteration loop · Test generation and execution · Error diagnosis from tracebacks · Multi-file project management
Three capstone projects combining everything into portfolio-worthy applications, plus comprehensive interview preparation: 60+ questions, system design, and coding challenges.
Multi-agent research team · RAG with hybrid search · MCP tool integration · Streaming API with Docker deployment
Domain data collection and preparation · QLoRA + Unsloth training · DPO alignment · Quantization and multi-format deployment
User authentication and rate limiting · Document upload with RAG · Agent with tools and memory · Cost tracking and production monitoring
Tokenization and embeddings Q&A · Transformer architecture questions · Training pipeline (SFT, RLHF, DPO) · Inference optimization and model comparison
RAG architecture and pipeline questions · Embedding and vector store Q&A · Advanced RAG patterns (CRAG, Self-RAG, GraphRAG) · Scenario-based RAG system design
Agent architecture and tool use Q&A · Memory and state management questions · Multi-agent pattern comparison · Production agent deployment scenarios
ChatGPT clone system design · Enterprise document Q&A platform · AI customer support system · Content generation pipeline at scale
Timed RAG pipeline implementation · ReAct agent from scratch · Semantic caching with Redis · MCP server coding challenge
Bias detection in LLMs · Fairness testing for differential treatment · Environmental impact of AI · Responsible AI guidelines
Portfolio and GitHub profile optimization · Technical writing for career growth · Resume optimization for GenAI roles · 4-week interview preparation plan