AI-native developer
A practical route from model fundamentals to grounded agents, reliable tool use, evaluation, security, and local deployment.
26 topics 5 checkpoints about 40 h beginner → advanced
0 of 26 topics · milestone 1 of 5 Continue: LLM application basics
The map
done you are here needs the checkpoint before it checkpoint
Milestone 1 · Model and prompt foundations
0 of 5 topics · about 64 min now LLM application basics 17 min 2 Natural language processing 17 min
soon Deep learning basics 10 min
soon Transformer architecture 10 min
soon Prompt engineering 10 min
Checkpoint 1 unlocks M2
Milestone 2 · Reliable model interfaces
0 of 5 topics · about 65 min 1 OpenAI API 18 min 2 Claude API 17 min
soon Structured output 10 min
soon Function calling 10 min
soon Mcp 10 min
Checkpoint 2 unlocks M3
Milestone 3 · Grounded knowledge systems
0 of 5 topics · about 54 min soon Embeddings 10 min
soon Vector database 10 min
soon Rag retrieval augmented generation 10 min
soon Knowledge graph 10 min
5 LangChain 14 min Checkpoint 3 unlocks M4
Milestone 4 · Stateful and multimodal agents
0 of 5 topics · about 50 min soon Ai agents 10 min
soon Langgraph 10 min
soon Multimodal ai 10 min
soon Vlm 10 min
soon Claude computer use 10 min
Checkpoint 4 unlocks M5
Milestone 5 · Production and adaptation
0 of 6 topics · about 74 min soon Llm evaluation 10 min
soon Ai observability 10 min
soon Llm jailbreak prevention 10 min
4 Hugging Face 17 min 5 Local large language models 17 min soon Llm finetuning 10 min
Checkpoint 5 done
Why this order: Start with language, neural-network, transformer, and prompt mechanics so model behavior is not a black box. Next, make provider responses and tool calls explicit contracts before adding retrieval. Build stateful and multimodal agents only after those boundaries are clear, then finish with evaluation, observability, security, reproducible artifacts, local serving, and adaptation.