AI AI SkillForge

40-hour project-based training plan

AI Fundamentals to RAG - Complete Training Plan

A 10-session engineering curriculum designed around three questions in every module: why do we need this, how does it work internally, and what will we build?

Enroll Now Back to Learning Paths

Target Audience

  • Software Engineers
  • Software Architects
  • QA Engineers
  • DevOps Engineers
  • Technical Leads
  • AI Enthusiasts

Learning Outcome

  • Understand how LLMs work internally
  • Write production-quality prompts
  • Build semantic search systems
  • Create vector databases
  • Build RAG applications
  • Optimize retrieval quality
  • Understand fine-tuning
  • Deploy an AI application

Technology stack

Tools used across labs, assignments, and the capstone project.

ProgrammingPython
IDEVS Code
NotebookJupyter
LLMOpenAI, Gemini, Local Llama
Embedding ModelSentence Transformers
Vector DBChromaDB, FAISS
FrameworkLangChain
FrameworkLlamaIndex
APIFastAPI
UIStreamlit
DatabaseSQLite
GitGitHub
ContainerDocker

Modules

10 sessions, 4 hours each.

Module 1 - 4 Hours

AI Fundamentals

Evolution of AI, AI landscape, transformer revolution, and AI use cases across healthcare, finance, retail, education, and manufacturing.

Demo: ChatGPT vs Traditional Search Lab: Install Python, VS Code, Jupyter, and OpenAI SDK Assignment: Create your first AI application
Module 2 - 4 Hours

LLM Basics

Tokens, BPE, embedding space, vectors, dimensions, similarity, LLM architecture, and model comparison.

Labs: Token counter, plot embeddings, call multiple LLM APIs Assignment: Compare outputs from GPT, Gemini, Claude, Llama, Mistral, and DeepSeek
Module 3 - 4 Hours

Prompt Engineering

Prompt anatomy, zero-shot, one-shot, few-shot, Chain of Thought, Tree of Thought, ReAct, role prompts, XML, JSON, and structured prompting.

Labs: Prompt competition and improve poor prompts Mini Project: Restaurant Chatbot
Module 4 - 4 Hours

Embeddings

Dense vectors, sparse vectors, cosine similarity, Euclidean distance, dot product, and embedding model selection.

Labs: Generate embeddings, compare similarity, semantic search Project: FAQ Search Engine
Module 5 - 4 Hours

Vector Databases

Why SQL fails for semantic search, approximate nearest neighbor, indexing, HNSW, IVF, FAISS, ChromaDB, Pinecone, Weaviate, Milvus, and Qdrant.

Labs: Store embeddings and retrieve documents Mini Project: Document Search Engine
Module 6 - 4 Hours

Retrieval Augmented Generation

RAG architecture, retriever, embeddings, vector DB, prompt builder, generator, chunking, sliding window, metadata, hybrid search, BM25, and semantic search.

Labs: Build RAG, PDF QA, website QA Mini Project: Chat with PDF
Module 7 - 4 Hours

Advanced RAG

Parent-child retrieval, multi-query retrieval, context compression, re-ranking, query expansion, Graph RAG, and Agentic RAG.

Labs: Improve retrieval quality and measure retrieval Project: Enterprise Knowledge Assistant
Module 8 - 4 Hours

Fine-Tuning Concepts

Prompt engineering vs RAG vs fine-tuning, when not to fine-tune, LoRA, QLoRA, PEFT, instruction tuning, supervised fine-tuning, RLHF, and DPO.

Lab: Fine-tune a small model Case Studies: Medical chatbot and legal chatbot
Module 9 - 4 Hours

End-to-End AI Application

Architecture, backend, vector DB, prompt layer, authentication, logging, caching, streaming, and deployment.

Lab: Build a production-style AI application
Module 10 - 4 Hours

Capstone Project

Students build one production-ready application such as HR resume screening, AI tutor, medical knowledge assistant, company policy bot, legal assistant, PDF chat, AI travel planner, research assistant, enterprise search, or customer support.

Outcome: Final presentation with source code, documentation, diagrams, tests, benchmark report, and deployment guide

Final deliverables

What every student submits at the end.

Source code on GitHub API documentation Architecture diagram Prompt design document Vector database schema RAG workflow diagram Docker deployment guide Unit tests Performance benchmark report 10-minute project presentation

Teaching pattern

Every session balances architecture, live coding, lab work, and review.

  1. 0-20 min: Business problem and learning objectives
  2. 20-60 min: Concepts with architecture diagrams and animations
  3. 60-90 min: Live coding from scratch
  4. 90-105 min: Break
  5. 105-165 min: Hands-on lab where students build the feature
  6. 165-210 min: Debugging, optimization, and best practices
  7. 210-240 min: Assignment briefing, Q&A, and recap