LLM Foundations 2.0 Path

Build a current, local-first foundation for model behavior, retrieval, evaluation, security, context, and cost.

flowchart LR
  M[Model mechanics] --> P[Prompt and context]
  P --> R[Retrieval]
  R --> E[Evaluation]
  E --> S[Security]
  S --> C[Cost and routing]

Lab	Title	Level	Time
Lab 004	How LLMs Work: Tokens, Attention, Training, and Hallucination	L50	45 min
Lab 005	Prompting and Context Engineering	L50	60 min
Lab 007	Embeddings and Local Similarity Search	L50	60 min
Lab 017	Reliable Structured Output with Local Validation	L100	60 min
Lab 018	Tool Calling with Typed Local Contracts	L100	60 min
Lab 019	Streaming Responses: Events, Cancellation, and Safe UX	L100	50 min
Lab 022	Local RAG: Chunking → Embedding → Search → Answer	L200	90 min
Lab 097	Interactive: Tokens, Sampling, and Context Windows	L50	30 min
Lab 098	Context Engineering for Agents	L200	75 min
Lab 099	LLM Evaluations 101: Golden Sets, LLM-as-Judge, and CI Regression	L200	90 min
Lab 100	OWASP Top 10 for LLM and Agentic Applications: Threat Modeling	L200	75 min
Lab 101	Token Economics: Cost, Cache, Batch, and Routing	L200	60 min

Free-first route

All required exercises run locally with deterministic fixtures. Hosted embedding, generation, evaluation, and observability services are optional adapters.

Phase 3 additions

Lab	Focus
102	Prompting vs RAG vs LoRA
103	Mini-GPT on CPU
042	Enterprise RAG and evaluations