Real-World Projects

Hands-on projects that cement every concept from the learning modules. All code lives on GitHub, so you can clone, run, and modify everything.

gaurav-pahuja-ai/ai-real-world-projects
RAG
Beginner

Semantic Vector Bot

A production-ready Q&A system that ingests PDFs, indexes them in ChromaDB, and answers questions with source citations. Powered by Google Gemini (free tier) with a Gradio UI.

Overlapping character-level chunking
Pure semantic search via ChromaDB
Local ONNX embeddings - no API needed
Streaming responses with source citations
PythonChromaDBGeminiGradio
RAG
Beginner

BM25 Keyword Bot

A RAG pipeline that retrieves context using pure keyword search - BM25 and TF-IDF - with zero vector databases or embeddings. Proves you can build accurate, fast Q&A without the overhead of a vector store.

No vector DB or embeddings needed
BM25 + TF-IDF keyword retrieval
Side-by-side accuracy comparison vs vector RAG
Ideal for small, structured corpora
PythonBM25TF-IDFrank-bm25Gemini
RAG
Advanced

Enterprise Knowledge Base Search

Multi-tenant RAG system for internal company knowledge. Supports Confluence, Notion, and Google Drive connectors with access-control-aware retrieval.

Multi-source document ingestion
Per-user access-aware retrieval
Re-ranking with Cohere
REST API with auth
PythonQdrantFastAPILangChainOAuth2
MCP
Beginner

Hello MCP Server

The smallest useful MCP server: one tool, no API keys, no external services. The exact code walked through in the Build Your First MCP Server lesson, under 150 lines start to finish.

One tool: analyze_text
Zero API keys or external services
The exact code from the mcp-build lesson
Connects to Claude Desktop and Cursor
PythonMCP SDK
MCP
Intermediate

GitHub MCP Server

A fully-featured MCP server that exposes GitHub operations (issues, PRs, file reads, search) as tools for any MCP-compatible AI model.

Tools: read_file, list_issues, create_issue
Resources: repo structure, README
Secure OAuth token handling
stdio and SSE transport
PythonMCP SDKGitHub APIasyncio
MCP
Intermediate

Postgres MCP Server

An MCP server that lets AI models query a Postgres database safely. Includes read-only enforcement, query validation, and schema discovery.

Read-only query enforcement
Schema discovery as resource
Query explain before execution
Row-level security support
PythonMCP SDKPostgreSQLasyncpg
Agentic AI
Advanced

Autonomous Research Agent

A ReAct agent that takes a research question, searches the web, reads papers, cross-references sources, and produces a structured Markdown summary on its own.

Multi-step ReAct loop
Web search + PDF reading tools
Source deduplication & ranking
Structured Markdown output
PythonOpenAITavilyReActLangGraph
Agentic AI
Advanced

AI Code Review Agent

A multi-agent system that reviews pull requests. One agent checks security, another checks performance, a third reviews code style, and an orchestrator combines all the findings.

Orchestrator-worker pattern
Parallel specialised agents
GitHub PR integration
Human-approval gate
PythonOpenAIGitHub APIMulti-agentLangGraph