Build a RAG App on Your Own Data | Hands-on 1:1 with Praveen Kumar

Want to understand RAG properly instead of copying another “chat with PDF” tutorial?

In this private 2-hour hands-on session, we'll build a complete Retrieval-Augmented Generation pipeline together.

We'll focus only on the foundations required to make a working RAG application.

No MCP.

No multi-agent architecture.

No advanced evaluation framework.

Those deserve separate sessions.

What we'll build

We'll implement this pipeline:

Documents

1. preprocessing
2. chunking
3. embeddings
4. vector storage
5. similarity search
6. context retrieval
7. LLM prompt
8. grounded response

By the end of the session, you'll understand what each stage is doing and why it exists.

What we'll cover

1. What RAG is and what problem it solves
2. When RAG should and should not be used
3. Loading documents
4. Cleaning and preprocessing data
5. Chunking documents
6. Chunk size and overlap
7. Creating embeddings
8. Understanding vector representations
9. Storing embeddings in a vector database
10. Similarity search
11. Retrieving relevant chunks
12. Passing context to an LLM
13. Building a grounded prompt
14. Returning answers from retrieved context
15. Basic citations / source references
16. Common reasons beginner RAG systems fail

What we'll build

A working application where you can ask questions against your own documents.

Example:

User question

→ embed query

→ search vector database

→ retrieve relevant chunks

→ inject context into prompt

→ call LLM

→ generate grounded answer

Example use cases

You can bring:

1. PDFs
2. technical documentation
3. study notes
4. product documentation
5. company knowledge-base samples
6. Markdown files
7. articles
8. project documentation

If you don't have data ready, I'll provide a practical dataset.

What you'll leave with

✓ A working RAG application

✓ Source code from the session

✓ Understanding of document chunking

✓ Understanding of embeddings

✓ Working vector search

✓ Retrieval + LLM integration

✓ Basic source attribution

✓ Clear architecture diagram

✓ Understanding of what to improve next

Who should book this?

Good fit if you are:

1. a developer getting started with RAG
2. an AI/ML learner building LLM applications
3. a backend engineer moving into AI engineering
4. building a portfolio project
5. confused about embeddings/vector databases
6. able to follow basic Python code

Prerequisites

You should know basic Python.

You do NOT need previous knowledge of:

1. RAG
2. vector databases
3. embeddings
4. LangChain
5. LangGraph

We'll build the concepts progressively.

What this session does NOT cover

To keep the session useful and realistic within two hours, we will not try to squeeze in:

1. advanced RAG
2. hybrid search
3. reranking
4. query rewriting
5. RAG evaluation frameworks
6. AI agents
7. MCP
8. multi-agent systems
9. production-scale architecture

Those are separate follow-up sessions.

Session style

This is a build-along session.

We'll explain the concepts while implementing them.

You're encouraged to:

1. code along
2. ask questions
3. share your screen
4. bring your own dataset
5. bring an existing RAG attempt if you're stuck

The goal is not to cover the maximum number of buzzwords. The goal is to leave with one working RAG system that you actually understand.