📅 Build your plan

Pick a pace and we distribute every remaining roadmap resource across your week and generate a calendar you can import anywhere.

1. Set your pace

Hours per week:

75

resources to cover

48h

timed content

7

weeks needed

At 8h/week you finish around Sunday, October 4, 2026. 🏁

Your week-by-week schedule

Week 1 (Aug 17 – Aug 23)

1.1h
  • Aug 23Big-picture: what an LLM is · P1 · 8m
  • Aug 23Full forward pass, embeddings, softmax · P1 · 27m
  • Aug 23Attention: Q/K/V, masking, multi-head · P1 · 26m
  • Aug 23How facts live in MLP layers (often skipped) · P1 · 8m

Week 2 (Aug 24 – Aug 30)

8h
  • Aug 24How facts live in MLP layers (often skipped) · P1 · 15m
  • Aug 24Tokenization/BPE + positional encoding · P1 · 53m
  • Aug 25Tokenization/BPE + positional encoding · P1 · 63m
  • Aug 25APIs, prompts, streaming, multi-provider · P2 · 6m
  • Aug 26APIs, prompts, streaming, multi-provider · P2 · 14m
  • Aug 26Structured output (JSON/Pydantic) · P2 · 20m
  • Aug 26Why RAG; embeddings; vector DB; cosine · P3 · 34m
  • Aug 27Why RAG; embeddings; vector DB; cosine · P3 · 4m
  • Aug 27Build a real RAG app · P3 · 20m
  • Aug 27Hybrid search: BM25 + dense + RRF · P3 · 45m
  • Aug 28Hybrid search: BM25 + dense + RRF · P3 · 69m
  • Aug 29Hybrid search: BM25 + dense + RRF · P3 · 69m
  • Aug 30Hybrid search: BM25 + dense + RRF · P3 · 69m

Week 3 (Aug 31 – Sep 6)

8h
  • Aug 31Hybrid search: BM25 + dense + RRF · P3 · 69m
  • Sep 1Hybrid search: BM25 + dense + RRF · P3 · 69m
  • Sep 2Hybrid search: BM25 + dense + RRF · P3 · 69m
  • Sep 3Hybrid search: BM25 + dense + RRF · P3 · 2m
  • Sep 3Reranking (cross-encoders/ColBERT) · P3 · 9m
  • Sep 3Query transforms (HyDE, multi-query) · P3 · 20m
  • Sep 3Chunking: parent-child, sentence-window · P3 · 20m
  • Sep 3GraphRAG / RAPTOR / Contextual Retrieval · P3 · 18m
  • Sep 4GraphRAG / RAPTOR / Contextual Retrieval · P3 · 2m
  • Sep 4Metrics: faithfulness, context precision/recall · P3 · 12m
  • Sep 4RAGAS hands-on in Python · P3 · 6m
  • Sep 4The one rule: fine-tune = behavior, not knowledge · P4 · 20m
  • Sep 4LoRA/QLoRA + rank, alpha, dropout, 4-bit NF4 · P4 · 15m
  • Sep 4Quick PEFT overview (LoRA/QLoRA/DoRA) · P4 · 4m
  • Sep 4End-to-end deep dive + code · P4 · 9m
  • Sep 5End-to-end deep dive + code · P4 · 69m
  • Sep 6End-to-end deep dive + code · P4 · 9m
  • Sep 6Agent vs workflow; ReAct loop · P5 · 10m
  • Sep 6ReAct in practice · P5 · 6m
  • Sep 64 patterns + evals (whole course) · P5 · 44m

Week 4 (Sep 7 – Sep 13)

8h
  • Sep 74 patterns + evals (whole course) · P5 · 69m
  • Sep 84 patterns + evals (whole course) · P5 · 69m
  • Sep 94 patterns + evals (whole course) · P5 · 69m
  • Sep 104 patterns + evals (whole course) · P5 · 69m
  • Sep 114 patterns + evals (whole course) · P5 · 69m
  • Sep 124 patterns + evals (whole course) · P5 · 69m
  • Sep 134 patterns + evals (whole course) · P5 · 10m
  • Sep 13Reflection / Tool use / Planning / Multi-agent · P5 · 20m
  • Sep 13Build 8 agent projects · P5 · 20m
  • Sep 13Context window as scarce RAM; write/select/compress/isolate · P6 · 14m
  • Sep 13Anthropic's mental model + failure modes · P6 · 5m

Week 5 (Sep 14 – Sep 20)

8h
  • Sep 14Anthropic's mental model + failure modes · P6 · 15m
  • Sep 14Short vs long vs episodic memory · P6 · 22m
  • Sep 14Memory taxonomy + tools (Mem0, Letta, Zep) · P6 · 8m
  • Sep 14LangChain vs LangGraph vs LangSmith · P7 · 12m
  • Sep 14LangGraph: state, nodes, edges, memory, HITL · P7 · 11m
  • Sep 15LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 16LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 17LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 18LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 19LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 20LangGraph: state, nodes, edges, memory, HITL · P7 · 69m

Week 6 (Sep 21 – Sep 27)

8h
  • Sep 21LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 22LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 23LangGraph: state, nodes, edges, memory, HITL · P7 · 69m
  • Sep 24LangGraph: state, nodes, edges, memory, HITL · P7 · 32m
  • Sep 24LangGraph full free course · P7 · 37m
  • Sep 25LangGraph full free course · P7 · 69m
  • Sep 26LangGraph full free course · P7 · 69m
  • Sep 27LangGraph full free course · P7 · 6m
  • Sep 27CrewAI / OpenAI SDK / AutoGen / ADK · P7 · 20m
  • Sep 27Tracing & debugging (LangSmith) · P7 · 20m
  • Sep 27Why MCP (2-min mental model) · P8 · 8m
  • Sep 27Why → Architecture → Lifecycle → Build · P8 · 15m

Week 7 (Sep 28 – Oct 4)

6.9h
  • Sep 28Why → Architecture → Lifecycle → Build · P8 · 5m
  • Sep 28Full build: server+client, tools, resources, prompts, SAMPLING · P8 · 63m
  • Sep 29Full build: server+client, tools, resources, prompts, SAMPLING · P8 · 12m
  • Sep 29Production server w/ FastMCP + auth · P8 · 57m
  • Sep 30Production server w/ FastMCP + auth · P8 · 45m
  • Sep 30Hands-on labs (multi-language) · P8 · 20m
  • Sep 30Tool poisoning, line jumping, rug pulls · P8 · 3m
  • Oct 1Tool poisoning, line jumping, rug pulls · P8 · 17m
  • Oct 1Full threat model + defenses · P8 · 20m
  • Oct 1Auth, credential risk, CVEs · P8 · 20m
  • Oct 1Objective vs LLM-judge; error analysis; traces · P9 · 12m
  • Oct 2Objective vs LLM-judge; error analysis; traces · P9 · 8m
  • Oct 2RAG evals (RAGAS): faithfulness, precision, recall · P9 · 12m
  • Oct 2RAGAS in code (run in CI) · P9 · 20m
  • Oct 2Tracing/observability in practice · P9 · 20m
  • Oct 2All 10 risks in 4 minutes (overview) · P10 · 4m
  • Oct 2LLM01 Prompt Injection (direct/indirect) · P10 · 4m
  • Oct 3LLM01 Prompt Injection (direct/indirect) · P10 · 21m
  • Oct 3Deep dive w/ real exploits & fixes · P10 · 30m
  • Oct 3Practitioner guide (every entry + controls) · P10 · 18m
  • Oct 4Practitioner guide (every entry + controls) · P10 · 0m