https://www.louisbouchard.ai/from-zero-to-hero-with-llms/
Goal: Design, build, secure, and operate production-grade RAG systems integrating enterprise data, using vector DBs, and applying adapters as needed.
Outcome: Understand LLM fundamentals, embeddings, chunking, RAG architecture, and prompt basics.
- Coursera Module: Generative AI with LLMs — Modules 1–2 — https://www.coursera.org/learn/generative-ai-with-llms
- OpenAI (RAG intro): Retrieval Augmented Generation & Semantic Search — https://help.openai.com/en/articles/8868588-retrieval-augmented-generation-rag-and-semantic-search-for-gpts
- Microsoft Learn (Azure): RAG in Azure AI Search (overview) — https://learn.microsoft.com/en-us/azure/search/retrieval-augmented-generation-overview
- YouTube (intro): LangChain RAG Tutorial for Beginners — https://www.youtube.com/watch?v=SQCtfJohQcE
- Build Your First RAG Application Using LlamaIndex!: — https://www.youtube.com/watch?v=krj6HbIrDdE&t=4s
Outcome: Build local RAG PoC: extract, chunk, embed, vector index (FAISS/Chroma), LLM prompt.
- Coursera Module: Module 3 – Application patterns: RAG pipelines
- OpenAI: Embeddings guide & RAG Cookbook — https://platform.openai.com/docs/guides/embeddings
- Microsoft Learn: Quickstart – Generative Search (RAG) with Azure — https://learn.microsoft.com/en-us/azure/search/search-get-started-rag
- YouTube (PoC): RAG + LangChain Python Project – https://www.youtube.com/watch?v=tcqEUSNCn8I
- YouTube (deep dive): LangChain – Chain Deep Dive (Map‑Reduce, Stuff, Refine…) — https://www.youtube.com/watch?v=OTL4CvDFlro
Outcome: Deep dive into vector DBs, embeddings storage & retrieval, similarity search, and semantic filtering.
- OpenAI: Vector DB connectors & examples (Pinecone, FAISS, etc.) — https://cookbook.openai.com/examples/vector_databases/pinecone/using_vision_modality_for_rag_with_pinecone
- Coursera: Supplement with LangChain & Vector DB videos or relevant modules in "Fundamentals of AI Agents Using RAG and LangChain" — https://www.coursera.org/learn/fundamentals-of-ai-agents-using-rag-and-langchain
- Microsoft Learn: Azure Cognitive Search vector search overview — https://learn.microsoft.com/en-us/azure/search/vector-search-overview
- YouTube: Vector Search Tutorial with Pinecone + LangChain — https://www.youtube.com/watch?v=LRy4bq1hhYE
- YouTube: Semantic Search with Azure Cognitive Search — https://www.youtube.com/watch?v=UL_bLP81dOY
Outcome: Replace local index, add metadata filtering, deploy to cloud container.
- Coursera Module: Module 3 continued (retrievers, filtering, re-ranking)
- OpenAI: Vector DB connector examples (continued) — https://cookbook.openai.com/examples/vector_databases/pinecone/using_vision_modality_for_rag_with_pinecone
- Microsoft Learn: Azure AI Search & RAG patterns — https://learn.microsoft.com/en-us/azure/search/retrieval-augmented-generation-overview
- YouTube (Azure demo): Cognitive Search + OpenAI integration — https://www.youtube.com/watch?v=9PmfonSzjlw
Outcome: Implement map‑reduce summarization, re‑ranking, conversational memory.
- Coursera Module: Module 4 – Summarization & memory pipelines
- OpenAI: RAG & multi‑stage summarization optimizations — https://platform.openai.com/docs/guides/optimizing-llm-accuracy/retrieval-augmented-generation-rag
- Microsoft Learn: Conversational AI & memory with Azure Search/OpenAI — https://learn.microsoft.com/en-us/shows/ai-show/transform-rag-and-search-with-azure-ai-document-intelligence
- YouTube (summarization): Summarization with LangChain — https://www.youtube.com/watch?v=w6wOhSThnoo
- YouTube (another method): Summarize Long Texts with LangChain – Map‑Reduce & Refine — https://www.youtube.com/watch?v=FBvnzcbGtfU
Outcome: Fine‑tune or use adapters on domain corpus; compare with RAG.
- Coursera Module: Module 4 conclusion – Fine‑tuning & adaptation
- OpenAI: Fine‑tuning guide & best practices — https://platform.openai.com/docs/guides/fine-tuning
- Microsoft Learn: Azure AI Foundry – RAG customization patterns — https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/retrieval-augmented-generation
- YouTube (LoRA tutorial): Fine‑tune with LoRA — https://www.youtube.com/watch?v=8N9L-XK1eEU
- YouTube (deep LoRA): Mastering LoRA: Efficient Fine Tuning for LLMs — https://www.youtube.com/watch?v=zVdrEkoM5Kk
Outcome: Private endpoints, Key Vault, RBAC, audit logs, CI/CD, hallucination tracking.
- OpenAI: Retrieval/attribution & responsible-use recipes — https://platform.openai.com/docs/guides/retrieval
- Coursera: MLOps & productionization with Azure ML — https://www.coursera.org/specializations/mlops-machine-learning-duke
- Microsoft Learn: Azure data privacy & HIPAA for OpenAI — https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy
- YouTube (demo): Enterprise GPT with Azure Cognitive Search + OpenAI — https://www.youtube.com/watch?v=A_gVmzAHEhU
Outcome: Autoscaling, batching, quantized inference, canaries, SLOs.
- Coursera: Revisit Module 4 or MLOps capstone deployment labs
- OpenAI: Model optimization & cost control — https://platform.openai.com/docs/guides/model-optimization
- Microsoft / Infra: NVIDIA Triton & HF TGI docs — https://docs.nvidia.com/deeplearning/triton-inference-server/user-guide/docs/index.html
- YouTube (inference): Getting Started with NVIDIA Triton — https://www.youtube.com/watch?v=NQDtfSi5QF4
- Retrieval precision@k, MRR
- Groundedness / hallucination rate
- Latency (P95)
- Token cost per query & per 1k queries
- Reindex/update turnaround time
- Compliance/PHI audit readiness
- Option A: I can generate this as a downloadable
README.md - Option B: I can draft a 12-week day-by-day study calendar with tasks from this roadmap
Which would you prefer— A or B?