Mini Courses.

Short, practical slide decks on building production GenAI systems. Free to download and use for learning or team sessions.

Course 01 · 11 slides

Agentic RAG: From Retrieval to Reasoning

Design retrieval systems that plan, route, check their own evidence, and cite sources.

For mL engineers and architects who already know basic RAG

  • Naive vs. agentic RAG
  • The agentic RAG loop
  • CRAG & Self-RAG
  • Hybrid retrieval
  • Multi-agent RAG
  • RAG evaluation
Course 02 · 11 slides

Storage Selection for AI Solutions

Choose vector, relational, graph, cache and object storage for real workloads.

For architects and engineering leads designing GenAI platforms

  • Decision framework
  • Vector DBs compared
  • pgvector vs. dedicated stores
  • HNSW, IVF, PQ, DiskANN
  • Agent memory & caching
  • Compliance
Course 03 · 11 slides

Agentic Frameworks Explained: CrewAI, LangGraph, AutoGen & More

How the major agent frameworks think, where each one fits, and how to choose.

For engineers choosing a stack for multi-agent systems

  • CrewAI
  • LangGraph
  • AutoGen / Microsoft Agent Framework
  • OpenAI Agents SDK
  • Google ADK
  • Pydantic AI
Course 04 · 11 slides

MCP: Model Context Protocol for Tool-Using Agents

The open standard for connecting LLM apps to tools and data: architecture, build, and security.

For engineers integrating agents with enterprise systems

  • Hosts, clients & servers
  • Tools, resources & prompts
  • Build a server in Python
  • MCP vs. A2A
  • Security model
  • Enterprise rollout
Course 05 · 11 slides

LLM Evaluation & Guardrails for Production

Measure quality, catch regressions before release, and keep models safe in production.

For teams taking LLM features from demo to production

  • Layers of evaluation
  • Golden datasets
  • LLM-as-judge
  • Guardrails pipeline
  • CI evaluation gates
  • Observability
Course 06 · 11 slides

Fine-Tuning LLMs with LoRA & QLoRA

When to fine-tune, how parameter-efficient methods work, and how to ship adapters.

For mL engineers adapting open models to domain tasks

  • Prompt vs. RAG vs. fine-tune
  • SFT, PEFT, DPO
  • How LoRA works
  • QLoRA
  • Hyperparameters
  • Serving with vLLM