v0.1.0 — the full applied-LLM track

First tagged release. The repo now covers the applied-LLM stack end to end, framework-free, on the free Groq API.

What's inside

Teaching notebooks (00–11) — one skill at a time, raw APIs, evals as the spine:

• 00–01 Setup, prompting fundamentals, structured output, tool calling, streaming, context & caching
• 02 / 04 Evals I & II — golden sets, LLM-as-judge, regression-as-CI (the spine)
• 03 RAG — retrieve → hybrid/rerank → chunking → why-RAG-fails
• 05 Agents — loop from scratch, tool design, guardrails, MCP, Skills, harness engineering
• 06 Adaptation — fine-tune vs RAG vs prompt; LoRA (with a Colab-T4-verified appendix)
• 07 Security — prompt injection & the trust boundary
• 08 Operations — observability/LLMOps, reliability, MLflow experiment tracking & registry
• 09 / 10 Serving & inference performance (vLLM/TGI/Triton/TensorRT-LLM) and ML system design
• 11 Customer craft — scoping & discovery

Case studies & capstone (12) — three runnable, end-to-end case studies:

• A Customer-support assistant: scoping → deployed → debugged in production
• B Contract extraction: pipeline vs agent, settled with eval + cost
• C Red-team robustness benchmark: an attacker→target→judge PAIR loop reporting ASR

Plus the capstone brief for the deployed project that goes on your resume.

Verified

GPU appendices (06 LoRA, 09 vLLM) run end to end on a free Colab T4; 08/03 MLflow runs locally. Everything else is verified statically and runs on the free Groq API.

Feedback and run-reports welcome in Discussions.