AI consulting and training
I help companies work out where AI is needed , and then I build it.
Three years of a PhD in computer vision, then two as CTO of a startup that took that research inside companies that manufacture things. Now I work on my own.
PhD in computer vision at Sapienza · Published at CVPR, ICCV, IROS · Co-founder and CTO of Procedo from 2024 to 2026
The problem
The three things I get told most often
"I need this part."
An email arrives with a crooked photo. Before anyone can quote a price, someone has to work out which component it is, and it can take two people a couple of days.
"Only Bruno knows how to do it."
Forty years of the trade sitting in one person's head. Nobody has ever had the time to write it down, and in two years Bruno retires.
"We tried ChatGPT."
Someone in the office started using it on their own. It isn't clear with which documents, or whether the company agrees, and so far nobody has had the chance to talk about it.
How I work
How a job usually goes
Four steps. You can stop after any of them, and what has been done up to that point stays yours anyway.
I listen
I spend half a day at your company, together with the people who do the work every day. It's how I understand how you work, and how you get an idea of the way I think.
What comes out of it
A list of the places where time is lost, ordered by what they cost.
Diagnosis
Two or three weeks to look at the data you actually have, the systems already in use and who takes care of what. At the end I write what would be worth doing, in what order, and which roads I'd leave alone.
What comes out of it
A document that stays yours, and that you can show to whoever you like.
I build
I always start with the part that might not work, so that if something turns out to be impossible we find out early, while changing course still costs little.
What comes out of it
Something your people open on Monday morning and actually use.
I stay
Software nobody looks after stops being useful within a few months. If it's needed we stay in touch: maintenance, servers and the changes that come up along the way.
What comes out of it
A person who knows the system and who you can call when you need to.
The evidence
Some of the things I've built
Mechanical engineering · real client project
3D Parts Portal
The customer opens the 3D model of the machine they bought, clicks the part they need, and the request reaches the company already matched to that component's internal code.
Industry · in production
Procedo
Video shot on the shop floor becomes step-by-step procedures, and the people doing the work can ask questions of their own company's documentation. I was co-founder and CTO for two years; today I follow it as an advisor.
Live product · open to everyone
Spannum
A working-memory test built to the protocol published in the literature, with the sources cited and the limits stated. It's online and you can try it.
Training
It starts with people
Four hours, at your company or in a classroom, to show people what these tools can do and where their limits are. We work on concrete cases.
Since February 2025 the AI Act requires companies that use artificial intelligence tools to ensure a minimum level of training for their staff, and to be able to document it. You'll do that course anyway: it may as well be good for something.
4 hours
Introductory course
For the whole company, from the owner to the shop floor. What has changed in the last few years, what has stayed the same, and the concrete risks to your data.
8–12 hours
Workshop by function
Engineering office, purchasing, administration, sales. Each group works on its own documents, so what they learn can be used the next day.
Length agreed
Funded programme
Through accredited training bodies and interprofessional funds, when the company wants a long programme without committing its own cash.
I taught for four years at Sapienza as a teaching assistant, and I was an invited speaker at Ferrari S.p.A.'s Data Science Hub. Courses are organised directly, or through accredited training bodies and interprofessional funds. How a course gets organised →
Where what I know comes from
From research to production
A master's in data science, then a PhD in computer vision at Sapienza, with the work published at conferences where it gets examined thoroughly before it comes out. I was then CTO of an industrial startup for two years, which is the job where you find out how much of that research survives contact with a real company.
Every paper I published came out together with the code. It's the habit I still work with: technical decisions stay written down, and the software I hand over opens and reads.
2021–2024
PhD in computer vision, Sapienza — PINlab
CVPR · ICCV · IROS
The main conferences in the field
2024–2026
Co-founder and CTO of Procedo, an industrial startup
Where I work
Venice is my base
I live and work in Venice. Most of the companies I work with sit in the Veneto and Friuli regions, close enough that the first step can happen face to face: half a day on the shop floor, alongside the people who do that job every day.
After that, distance matters little. Diagnosis, building and maintenance all carry on remotely, and courses run in a classroom or online, whichever suits you better. If you are further away, we meet once and then keep going at a distance.
In person
Veneto and Friuli-Venezia Giulia
Remotely
The rest of Italy, and beyond
Courses
In a classroom or online