Generative AI Learning Path

A practitioner's checklist for getting started with GenAI development — the Python patterns, API skills, terminal comfort, and tooling you need before writing your first LLM application. No math required.

A practitioner's mental model of how LLMs work — what transformers do, why tokenization affects your costs, how embeddings power search, and what controls your model's behavior. No math required.

A practitioner's guide to choosing the right model — when to use large vs small, closed vs open, text-only vs multimodal. Includes cost comparisons, capability tradeoffs, and a decision framework for production model selection.

A practitioner's guide to calling LLM APIs in production — the chat completions interface, provider SDK patterns, streaming, cost management, multi-provider routing with LiteLLM, and structured outputs with Pydantic.

A practitioner's guide to fine-tuning LLMs — when to fine-tune vs RAG vs prompt, how LoRA and QLoRA work without the math, data preparation checklists, cost comparisons, and the OpenAI and HuggingFace workflows.

A practitioner's guide to self-hosting LLMs — when to self-host vs use APIs, how to choose between vLLM, TGI, and Ollama, GPU sizing, cost comparisons, and production deployment patterns with FastAPI, SageMaker, and Docker.

A practitioner's guide to production prompt engineering — system prompt design patterns, few-shot template libraries, chain-of-thought techniques, structured output extraction, and prompt security defenses. Patterns that work at scale, not theory.

A practitioner's guide to building RAG systems that work in production — when to use RAG, chunking strategy comparison, embedding model selection, vector database tradeoffs, retrieval quality tuning, and a production-readiness checklist.

A practitioner's guide to upgrading basic RAG — hybrid search, reranking, query transformation, parent-child chunks, self-corrective loops, and when each technique is worth the added complexity. Decision frameworks, not algorithms.

A practitioner's guide to LLM agents — ReAct patterns, tool definition best practices, when to use agents vs. chains, memory strategies, multi-agent architectures, human-in-the-loop gates, and production guardrails. Patterns that ship, not theory.

A practitioner's guide to evaluating LLM applications — what to measure, LLM-as-judge patterns, RAGAS for RAG pipelines, regression testing, A/B testing, evaluation pipeline design, and when to invest in human eval. Decision frameworks for building eval infrastructure that earns trust.

A practitioner's guide to building production guardrail systems — input validation, output filtering, prompt injection defense, PII detection, and framework selection. Defense in depth for every LLM application.

A practitioner's guide to MCP — what it solves, when to use it vs. direct tool calling, how to build servers with FastMCP, security considerations, and practical setup for Claude Desktop and agent workflows.

A practitioner's guide to deploying GenAI on AWS — Bedrock vs. SageMaker decision framework, serverless architectures, data service selection, security patterns, and cost optimization strategies.

A practitioner's guide to building AI workflows without code — when no-code makes sense, n8n workflow patterns for LLMs, connecting AI to business tools, limitations vs. custom code, and production deployment.

Build a production-grade document Q&A portal from scratch — document ingestion, chunking, embeddings, vector storage, RAG retrieval, FastAPI backend, and chat frontend. A practitioner's end-to-end project.

Build an autonomous multi-agent system that researches a topic, gathers data from multiple sources, synthesizes findings, and produces a structured report — a practitioner's end-to-end agentic project.