Hedi Manai
EN FR

Learn · Understand · Build

Hedi Manai AI Agent Builder.

I build AI agents that run in production — multi-agent systems, RAG pipelines, MCP servers. What I learn building them becomes a lesson in the Playbook; the code that proves itself becomes a tool in the Toolkit.

Start the Playbook Browse the toolkit

About

Engineering systems that run in the real world.

With a background spanning data science, machine learning, R&D leadership, and backend engineering, I bridge the gap between AI research and production reality. I architect autonomous multi-agent systems and agentic AI workflows that solve complex tasks through intelligent orchestration and coordination.

Using the Model Context Protocol (MCP) and frameworks like LangGraph and CrewAI, I build stateful, resilient systems that move beyond prototypes — from workflow automation and data pipelines to full production deployments, all running on scalable Python and FastAPI infrastructure.

Technical stack & expertise

Multi-Agent Orchestration · RAG & Vector Architectures · Data Science & Machine Learning · Production Backend · MCP & Tool Design · Goal-Driven Automation & RPA

More about me

The Agent Engineering Playbook

Mastering the art of agentic systems.

A structured curriculum for engineers building real AI systems. What works, what breaks, and why — no fluff, no theory for its own sake.

Prompt Engineering What Is Prompt Engineering? A First Principles Introduction A rigorous first-principles introduction to how prompts work, why they fail, and the core techniques that matter in practice. Read lesson Prompt Engineering Building the Production System Prompt Master the five-component anatomy and decision protocols required to build reliable system prompts for production agents. Read lesson Prompt Engineering Chain-of-Thought and Self-Consistency The technique that changed how we think about AI reasoning — the mechanism, seven variants, and self-consistency sampling. Read lesson

View all lessons

Toolkit

The toolkit. Install what I run.

What's here runs in my own stack first. Open source, one-command install, documented failure modes.

Claude Code Skill v1.3.0 Guardsman: No Diff Ships Unchallenged Reads your repo's conventions, sizes every change by its blast radius, and makes the check behind non-trivial logic actually run before code ships. MIT Updated View docs Python SDK v1.0.0 ToolOps: The Service Mesh for AI Tools Wraps any Python tool an agent calls with the resilience layer production systems expect from a service mesh — semantic caching, circuit breakers, retries, and OpenTelemetry traces, via a single decorator. Apache-2.0 Updated View docs

Browse the toolkit

Contact

Got a system to build?

If you're building a RAG pipeline, an AI agent workflow, or an MCP integration and want to compare notes, swap ideas, or just avoid the mistakes I've already made — reach out.

Email me

Or get the Playbook in your inbox — one email when a new lesson ships, and nothing else.

No noise, no spam. Unsubscribe anytime.