Stillwind Thinks Finding Electronic Parts Should Be as Easy as Texting

The Macro: Hardware Engineering Has a Search Problem

If you have ever tried to find an electronic component for a hardware project, you already know the pain. The major distributor catalogs from Digi-Key, Mouser, and Arrow are enormous. Millions of parts, each described by dozens of technical parameters, organized in taxonomies that make sense to the database but not necessarily to the engineer who just needs “a low-power microcontroller with I2C and at least 64KB of flash.”

The search interfaces on these platforms were designed in an era when filtering by parametric specifications was considered innovative. You select a category, then a subcategory, then start narrowing by voltage, package type, operating temperature range, and thirty other fields. It works if you already know exactly what you are looking for. It fails badly when you are in the exploratory phase of design, which is where most of the interesting engineering decisions happen.

This matters at scale because hardware development timelines are directly tied to component selection speed. Every hour an engineer spends hunting through parametric tables is an hour not spent on actual design work. And when supply chain disruptions hit, which they have done with alarming regularity since 2020, finding viable alternative components becomes even more critical and even more painful.

The electronic components market itself is enormous. Global semiconductor revenue alone exceeded $500 billion in recent years, and the broader passive and active components market adds hundreds of billions more. The tools that engineers use to navigate this market have not kept pace with the market’s growth or complexity.

Stillwind, backed by Y Combinator (W25), is starting with search and aiming at something much bigger: autonomous electrical engineering.

The Micro: Natural Language Meets Millions of Components

The founding team is four computer science graduates from ETH Zurich, which is about as strong a technical pedigree as you will find for this kind of problem. Linus Meierhoefer (CEO) comes from complexity theory research and quantitative engineering at DRW. Lukas Ego (CPO) worked on deep learning optimization. Hannes Furmans (CTO) contributed to open source work on libp2p. Josef Zoller (Chief Engineering Officer) brings systems, networks, and machine learning expertise. All four studied CS at ETH. That is a team built for hard technical problems.

The first product, Stillwind Search, converts natural language queries into detailed electronic part specifications and matches them against a proprietary database of millions of components. Instead of clicking through parametric filters, you describe what you need in plain language and the system translates that into the precise technical specifications required to find the right part.

This is harder than it sounds. Electronic components have deeply technical specifications that are not always described consistently across manufacturers. A capacitor from one manufacturer might list its ESR differently than the same class of capacitor from another. Mapping natural language to these specifications requires both strong NLP and deep domain knowledge of how electronic components are actually described, categorized, and cross-referenced.

The product is currently in a waitlist phase, which suggests they are being deliberate about the rollout. The website lays out a three-phase roadmap. Phase one is the semantic search layer. Phase two adds real-time simulation capabilities, including digital simulation, analog circuit modeling using machine learning, and firmware integration. Phase three introduces spatial reasoning for component placement and routing.

That roadmap is ambitious. If they execute on all three phases, they are not building a search tool. They are building an AI co-pilot for the entire hardware design process. Component selection, circuit verification, and PCB layout, handled by software that understands the constraints and trade-offs that currently require senior engineering judgment.

The competitive field here is interesting. Octopart, now owned by Altium, is the most well-known component search engine, and it is decent but fundamentally parametric. FindChips does aggregation across distributors. Neither uses natural language in any meaningful way. On the EDA side, Altium, Cadence, and KiCad all have component libraries, but their search is an afterthought within a larger design tool.

What Stillwind is doing is treating search as the entry point to a much larger platform play. That is a classic wedge strategy, and it works if the wedge is good enough that people actually use it.

The Verdict

I think the thesis here is strong and the team is exceptionally well-matched to the problem. Hardware engineering tooling is ripe for AI-driven improvement, and the fact that component search has barely changed in twenty years tells you there is real opportunity.

At 30 days, the question is whether the natural language search is accurate enough to be trusted. Engineers are precise people. If the tool returns wrong results, they will not come back.

At 60 days, I would want to know the size of the component database and how it compares to Octopart’s coverage. Breadth of coverage is table stakes for a search tool in this space.

At 90 days, the simulation roadmap either starts to materialize or it starts to look aspirational. The jump from “search” to “simulation” is enormous, and the credibility of the long-term vision depends on showing progress on phase two.

What would make this work is nailing the accuracy of search so thoroughly that hardware engineers start reaching for Stillwind before they open Digi-Key. What would make it fail is spreading too thin across the roadmap before the foundation is solid.

The ETH Zurich pedigree and the specificity of the problem give me confidence. This is the kind of company that either builds something deeply important or runs out of runway trying. I am betting on the former.