PCB Layout Finished 10x Faster with AI? Here’s How...

Robert Feranec
Robert FeranecMay 21, 2026

Why It Matters

Accelerating PCB layout by an order of magnitude shortens product development cycles, giving hardware firms a competitive edge without sacrificing design integrity.

Key Takeaways

  • Quilter focuses exclusively on accelerating PCB layout, not schematics.
  • Uses reinforcement learning on geometry, not LLMs, to optimize designs.
  • Targets 200‑1,000 component boards, cutting weeks to days.
  • Integrates with Altium, Cadence, Expedition, keeping engineer control.
  • Treats AI as a junior engineer, requiring human oversight.

Summary

The video introduces Quilter, a startup applying artificial intelligence to the PCB layout stage of hardware design, and explains how its founders aim to shrink the traditionally slow layout process.

Quilter deliberately avoids LLMs, treating layout as a geometry‑and‑physics problem solved with custom computational‑geometry engines and reinforcement‑learning algorithms. By training the model on design rules—DFM, DRC, signal and power integrity—rather than on billions of existing boards, the system iteratively proposes and evaluates layout variants until constraints are met.

Ben cites his SpaceX experience, noting layout as a major bottleneck, and likens the training method to DeepMind’s AlphaGo, where the AI learns by playing against itself. A live demo shows a simple Raspberry Pi‑based board being routed in Altium, illustrating how the tool acts like a junior engineer that suggests component placement and routing while the human retains final control.

If Quilter can reliably reduce a two‑to‑four‑week layout to a few days for 200‑1,000‑component boards, companies in aerospace, automotive and consumer electronics stand to accelerate time‑to‑market and lower labor costs, while still requiring engineer oversight to ensure quality.

Original Description

From estimated 10 weeks down to 1 week. Real examples showing how AI helps with component placement and PCB layout. Thanks to Sergiy Nesterenko and Ben Jordan from Quilter.
Links:
- Get started: https://app.quilter.ai
- Ask your questions about electronics here: https://fedevel.com/assistant
- Easy search through my videos: https://fedevel.com/hw-assistant
- Learn more about electronics, check out our online courses:
Chapters:
00:00 What is this video about
00:42 What makes Quilter different (focused only on layout)
02:09 Not an LLM, not ChatGPT: a geometry and physics problem
04:16 Inspired by DeepMind's AlphaGo and AlphaZero
08:57 Integrations: Altium, Cadence, Xpedition, KiCad
09:39 There is no magic button. Quilter is a tool, not magic
10:39 The Goldilocks zone: 200 to 1,000 components
14:05 Demo 1: a custom RP2350 microcontroller board in Altium
18:01 Uploading to the Quilter web app
32:52 Auto-detecting switching regulators and bypass caps from the schematic
44:48 Kicking off the placement job
52:35 Reviewing the first placement
53:57 Why floor planning is essential for high-quality results
1:32:16 Customer success criteria: "do I start from scratch, or start from Quilter?"
1:35:25 Demo 2: Project Speedrun, a full Linux-capable computer
1:37:49 Robert guesses how long the layout would take
1:40:07 The real quotes: 350 to 450 hours from professional contractors
1:41:13 Ben's one week of cleanup, fabricated at Sierra Circuits
1:44:33 Validation: Linux, Ethernet, USB, HDMI, Chrome, browser games, Doom
2:00:21 How to get started
2:02:06 About Quilter: 6 years, 35 people, $40M raised
------------------------------------------------------
(C) FEDEVEL by Robert Feranec

Comments

Want to join the conversation?

Loading comments...