▶ InfoCombiner
AI

Odesa-based AI agent designs PCBs: what it means for Ukrainian developers

Odesa-based startup 1UA-AGE has unveiled the first reproducible outcomes of its AI agent for PCB design: 100% routing on two challenging test circuits in KiCad. The team frames it as an autonomous "AI engineer" capable of taking over the routine part of electronics development.

The Ukrainian deeptech team 1UA-AGE from Odesa has reported the first reproducible results from internal testing of its AI agent for PCB design. On two benchmarks — the STRF radio-frequency circuit and a four-layer STM32 board — the autonomous system achieved 100% routing of connections in KiCad.

What 1UA-AGE actually built

The developers position their product as an "AI engineer": an agent that doesn't merely suggest circuit snippets but actually performs routing — one of the most labor-intensive parts of PCB design. Instead of an engineer manually laying every trace between components, the algorithm takes the task end to end, while the specialist reviews the output and makes adjustments.

Why STRF and STM32 are a serious test

STRF is a radio-frequency circuit sensitive to trace length and geometry: a routing mistake can effectively destroy the device's performance. The four-layer STM32 is a typical board for serial microcontroller devices with dense topology. Both benchmarks have long been used by engineers to evaluate the quality of autorouters in KiCad — they're the standard way to check whether a tool handles real-world, not "toy," problems.

Why this matters for small business

For small teams, PCB design is a bottleneck: a qualified engineer costs a lot, and manual routing of complex boards takes days. An autonomous agent that reliably delivers 100% connections on difficult circuits has the potential to shrink the timeline from prototype to pilot batch. That is critical for anyone building proprietary IoT devices, controllers, medical or industrial sensors — areas where every week of delay eats into margins.

Moreover, 100% routing is not a "perfect board"; it's just the first technical threshold. EMC, thermal behavior, manufacturability with a specific fabricator, and DFM rules still come next. So the market will be looking not only for "connections done," but also for metrics on signal quality, board cost, and the number of iterations before production.

Market context

KiCad remains the most popular free EDA tool among small teams in Europe and Ukraine — it's what most startups and contract developers build around. The emergence of AI agents for open-source EDA tools is a distinct trend that could lower the barrier to hardware development for companies without an in-house electronics department.

Bottom line

1UA-AGE has shown that a Ukrainian AI agent can match classical autorouters on atypical and demanding circuits. It's a technical signal that the tool is already useful for routine work. What the market needs next is independent testing, DFM rule support, and a clear access model for small teams.

💡 Need help with this topic? Learn about our service — AI Process Audit.

Author: Andrew Syromyatnikov · Founder of InfoCombiner

This article was drafted with AI assistance and reviewed by our editorial team. Editorial Policy