Product leader · Milan · AI products since 2017

Find clarityin chaos.

I turn new AI capabilities into products people can understand and trust. I also help teams change how they work.

I started building with AI in 2017. Aipoly gave it the sense of sight. Today, I turn human knowledge into AI. I stay close to product, design, engineering, and the people who use what we make.

Read on

Most product work starts in a mess. I find the useful part, give it a clear shape, and make it dependable.

For a decade, I have turned
new technology into practical products.

  1. 2015ForzaReal-time systemsA late answer is often a wrong answer.
  2. 2016Open researchI learned from the codeBuild it once. Then the paper makes sense.
  3. 2017AipolyGive AI the sense of sightSight had to be useful, fast, and clear.
  4. 2018 to 24V7 DarwinTurn visual knowledge into AIThe workflow is the product.
  5. 2023BenchLLMTests for LLM productsWrite the expected behaviour down.
  6. 2024 to nowV7 GoTurn human knowledge into AIShared context before more process.

Hover or focus a milestone

I work in a loop with
customers and agents.

Customer conversations set the direction. Claude helps me challenge the problem and design options. Codex turns decisions into working software. My personal AI operating system supplies context, tools, and checks. I review each result and take it back to customers.

Andrea's part

I keep the direction, taste, and accountability.

I decide which question matters, what good looks like, and what is ready to reach a customer.

  1. Customer01

    Say what hurts

    Can I review this before it runs?
  2. Andrea02

    Frame the decision

    What must be true?
  3. Claude03 ↗

    Challenge + design

    Which option best fits the work?
  4. Codex04 ↗

    Build + test

    Make the smallest useful change.
  5. Andrea05

    Review + ship

    Is this ready for a customer?
  6. Customer06

    Use + respond

    This works. That part still gets in the way.
Personal AI operating system

Shared context / tools / decisions / what we learned

Products that shaped my practice.

Six products and projects from the past decade, with the contribution I made and the lesson I kept.

Keyboard1–6 open previous / nextEnter toggleEsc close
01 / 2024 to nowV7 GoTurn human knowledge into AI

At V7, our mission is to turn human knowledge into AI. V7 Go lets experts teach agents how their work should be done.

I lead product across customer problems, interface design, agent behaviour, and evaluation. The deeper challenge is to give people and agents enough shared context to make good decisions and own the outcome.

Visit V7 Go
02 / 2023BenchLLMTest LLM products

We made BenchLLM to see whether an LLM product still worked after a model or prompt changed.

I made the open-source library with Simon Edwardsson in 2023. Teams could write down the behaviour they expected, compare models and prompts, and catch regressions in CI.

261GitHub starsMITLicensedPythonLibrary
Source on GitHub
03 / 2018 to 24V7 DarwinTurn visual knowledge into AI

I joined V7 as its first engineer and helped build Darwin from the first release.

Darwin helped computer vision teams turn visual knowledge into AI. It brought data, annotation, model-assisted workflows, and review into one product. My role grew from engineering into product direction as the platform and team matured.

Visit V7 Darwin
04 / 2017AipolyGive AI the sense of sight

Aipoly set out to give AI the sense of sight. We made that useful through a visual assistant for blind people.

I wrote computer vision models for a visual assistant on a phone. It had to recognise an object quickly, speak clearly, and earn trust. The model, the interface, and the person in front of the camera were all part of the same product.

Aipoly archive
05 / 2016 to 17Open visionTurn research papers into code
06 / 2015Forza FootballBackend systems for live scores

At Forza, I learned how much calm engineering matters to a live product.

I worked on the Elixir backend at what was then Football Addicts and is now Forza Football. The product handled millions of small, time-sensitive events. Fans felt the quality of the infrastructure every time a score arrived on time.

Visit Forza

Code led me
to product.

I’m Andrea Azzini, a product expert based in Milan. I work across product, design, engineering, and AI.

I help small teams turn difficult technology into clear, useful products. I stay close to the code and customers, ask simple questions, and care about how the finished thing feels.

Portrait of Andrea Azzini
Andrea AzziniProduct / AI / Teams