KiCon Europe 2025

This session is a practical demonstration of what is possible when AI is given the right tools to understand a hardware project. We will show how this structured approach moves beyond simple chatbots to create a genuinely useful design copilot.

The talk will be structured as follows:

1. Live Demo: An AI Copilot for KiCad (15 min)

We will begin with a live demonstration of the AmpereBrain AI assistant, showcasing its multi-level understanding of a KiCad project:

• High-Level Task (Project Specs): We'll ask the AI to "Update the specifications to require reverse-polarity protection on the main power input." The AI will read and apply a diff to the specifications.md file.
• Structural Analysis (Schematic View): We'll then ask, "Show me the power input circuit and check if it has a protection diode." The AI will generate a Mermaid diagram of the relevant schematic section for review.
• Detailed Query (Component Data): Following up, we'll ask, "What is the maximum reverse voltage of diode D1?" The AI will find the component, access its linked datasheet, and provide the specific parameter.
• Basic Design Review: Finally, we'll ask the AI to "Verify that every IC has a bypass capacitor connected between its power and ground pins," demonstrating its ability to traverse the netlist to perform simple design rule checks.

2. How It Works: A Structured Approach to AI (10 min)

After the demo, we'll briefly explain the core concepts that make this possible:

• The Foundation: Multi-Level Context: We'll show how the AI builds a holistic understanding by processing the project at three levels:
  1. Intent: Reading Markdown files (requirements.md, specifications.md) to grasp project goals.
  2. Structure: Parsing the schematic to create block diagrams and understand circuit topology.
  3. Detail: Accessing individual symbol properties and datasheet content.
• Organizing the AI Workforce: We'll introduce our two agent architectures: a hierarchical "Orchestrator" for delegating simple tasks and a collaborative "Swarm" for solving more complex problems. We'll outline the roles of the specialist agents (circuit expert, documentation writer, calculator).

3. The Future: From Proof-of-Concept to Product (5 min)

This proof-of-concept is the first step. We will conclude by discussing the roadmap to a commercial product:

• The Key to Reliable Editing: Schematics as Code. We'll explain our core principle: for an AI to safely edit a design, schematic elements like placement and wiring must be represented as structured, version-controllable text.
• The Roadmap: We will outline the path from this POC to a robust design tool. While the core file-parsing engine is open-source, the advanced AI features shown are part of the AmpereBrain commercial offering. Our goal is to build a reliable, professional tool that leverages AI to accelerate the hardware design process.