Meddle: AI and IoT put to work on production data

Meddle is our Industrial IoT platform: it collects the data factory machines produce every second and turns it into decisions. No hardware to buy, no months of setup: it connects to the machinery and you are running within minutes, through an interface where everything is configured by dragging blocks into place.

We enjoy telling this one because it is entirely ours, from the invention onwards: the idea, the market analysis, the design, the development, the business, the public funding and the communication. It is the kind of product that comes out of our startup studio when we decide to solve a problem we keep seeing in every manufacturing company: valuable data, scattered, unreadable and unused.

It did not start from a business plan, it started from a consulting job

Between the end of 2021 and 2022 we were working for a company doing data analysis with machine learning. Before analysing anything, though, that company had to solve a far more mundane problem: getting the data out of the plant. The machines were talking, but each one spoke a different language, and every integration turned into a project of its own.

The paradox we found in the field is this: industrial software for that job has existed for decades and is extremely powerful, but only a fraction of what it can do ever gets used. Configuring it takes a very specialised person, and companies often do not have one. And the people who actually need the data, meaning data scientists, analysts and production managers, have neither the time nor the inclination to become experts in fieldbus protocols.

So we built the first version of the platform. No sophisticated interface: an engine, and a set of small languages for defining the rules for data format and sampling. Spartan, but it did something the others did not. That is when we realised we were inside a niche nobody was serving.

Forward Factory: first a no, then 235,000 euro

We applied to Forward Factory, the digital manufacturing accelerator in CDP Venture Capital's national network, run by GELLIFY. The answer: no. The product was too embryonic, the market validation too thin.

They were right. Rather than filing the rejection away we used it as a to-do list: the user interface rethought, the commercial strategy made concrete, first real customers to validate against properly. Between the end of 2023 and 2024 we applied again, with more traction and a far sturdier product. This time it worked: 235,000 euro of funding and, in July 2024, the formal incorporation of the company.

Over half a million raised, public and private

From incorporation onwards we have run two tracks at once: growing the product and raising the money to fund it. Web Summit 2024, to test ourselves against a non-Italian market and start building investor relationships. B4i, Bocconi for Innovation: selection passed and 80,000 euro of private investment. Regional grants in Emilia-Romagna and Tuscany: 300,000 euro of public funding for product development, marketing and equipment.

It is also why public funding is one of the lines in our list of services: we learned it by doing it on ourselves, with our own money and our own deadlines.

Where we are today

Meddle is in production inside a company, not in a demo. In parallel we are building out the commercial side, with a dedicated sales hire.

Where Meddle is going: asking, not querying

Meddle exists to carry data from the physical world, robots, PLCs and sensors, into digital systems, and to make it readable without going through an integrator. The next step is removing the last technical rung as well: the analysis. We are adding AI features that let people ask questions in plain language, how much line 3 produced this week, or where the most time was lost to machine downtime, and get the answer already filtered, without writing a query and without building a dashboard.

Somebody who knows the factory should not have to become a data scientist to understand what is happening in their own factory.

That is the idea we started from in 2021, and it has not changed: what changes is how far down we can push the technical barrier between a question and its answer.