Engineering OS: Why Regulated Hardware Needs Its Own Operating System

A recap of Koddex founder Stéphane Dicostanzo's talk at the Tech for Industry Show. Watch the full pitch below, then read the recap.

Software was the easy part

Over the past year, AI has reshaped how software gets built. We moved from a chat window handing back snippets of code, to copilots embedded in the IDE, to autonomous agents that write software on our behalf. That shift is real, and it is only accelerating.

The same transformation is coming to industry. Anyone working in the sector can already feel it. But there is a catch: software was the easy version. Industry is a different story entirely.

The complexity problem

In industry, a single product is tens of thousands of components, thousands of technical specifications, and hundreds of regulatory standards. None of it stands alone. Everything is linked, and those links create dependencies, an almost infinite web of them. Change one measurement on one part, and you have to rerun an impact analysis across the entire system. The complexity compounds immediately.

To manage this, the current ecosystem relies on vertical, domain-specific tools. PLM, QMS, and ERP systems are powerful and highly specialized, but they are also data silos, and they are rigid. So to track what actually happens on the ground, teams fill the gaps with Excel files, Monday, Notion, and Jira, anything that stays close to the real process.

The result is that data gets scattered and duplicated. Nearly half of a project's cost ends up going not into the product itself, but into its structure: synchronizing data, holding sync meetings, and resolving the non-conformities that the fragmentation created in the first place.

This is hard for AI to operate in. But more importantly, it is hard for the teams doing the work. That is the problem Koddex set out to solve, for the teams first.

What Koddex is

Koddex is the foundation of your engineering system. It lets you model your business processes, model the data those processes require, and then instantiate that data inside an environment you have structured yourself.

Instead of creating records in Excel files or in Monday boards that anyone can reshape at will, you create them in a system that holds its structure. You connect your existing reference sources so the data stays clean, and then you carry that process across the entire company.

This matters because data does not stay in one place. A part defined by the design team is also needed by sales during handover, by maintenance, and by integration teams when the product gets installed. Continuity has to run end to end. That is what Koddex delivers.

Built for engineering

Because this is the engineering world, Koddex includes capabilities that generic software does not provide.

Traceability. Compliance is a genuine concern in industry. You need to know which requirement went into which component, to run which test, to validate which regulation. By building a knowledge graph, Koddex gives you, at any point in time, the full set of connected data.

Locking and revision. PLM users will recognize the concept, but the physical world changes the stakes. Once a product ships, it lives at the customer's site. A code deployment will not update it. The product has to keep living in its delivered form, while still being able to evolve, which means creating a new revision to move it forward.

Governance and security. A critical layer, treated as such.

Impact analysis. Thanks to global modeling, when you change a part, an attribute, or a formula, you can see exactly what it affects. Does it touch the sales process? The maintenance process? With Koddex, the affected elements light up instantly. It is efficient for people, and it is just as valuable for AI.

Good for humans, good for AI

AI loves data, but data alone is not enough. Koddex also gives the model context: the map of what is connected to what. Rather than churning through your entire dataset, the AI pulls only what is relevant. That saves tokens, and it cuts down on hallucinations.

Where Koddex stands today

Koddex is a year and a half old, in production, and live, with a proven onboarding system. We work with teams to map their business processes, help them build their first model, and then train them to evolve it autonomously inside their own organizations.

Deployments are measured in days, not years. This is not a two-year integration project. Koddex connects to your data sources and you can start iterating quickly. Just as we evolve the model in an agile way, you keep control to move incrementally. No big bang, no multi-year conversion effort.

We are currently being deployed by an industrial unicorn: more than a thousand people, across three continents, using Koddex to manage their entire engineering function. Working with a major industrial client has sharpened our approach to onboarding, structuring, and requirements. We are a young startup, but we have already been through real fire.

Who it is for

Koddex targets regulated industries: robotics, medtech, and aerospace. The common thread is regulation, combined with mechanics, electronics, and software in a single product.

Want to see what this looks like on your own processes? Explore the Engineering Operating System or book a demo.