Mountain View, California

Alessandro
Fin

Director of Products for Gemini Agents at Google DeepMind. Inventor of Google Lens and Pixel Face Unlock

I started my career writing programs that generate tests for hardware designs. Today I help build models that write the programs. Between the two: a PhD in Verona, engineering and leadership years in Dresden, Amsterdam and Silicon Valley, and Google.

Scroll to follow the thread

I Now

Gemini, at Google DeepMind

In 2025 I began a new chapter as Director of Products for Gemini Code Generation, Agents, and Thinking; after helping YouTube to use DeepMind video understanding technologies to help casual content creators. My current job is a professional and personal dream: a role that sits exactly where my interests meet — frontier models, agentic capabilities, and software development.

i. Code generation

Gemini as a software collaborator: reading a codebase, writing and editing it, explaining it, and reasoning about what a change will break. The goal is a model developers trust with real work, in the terminal, in the editor, and in the pipeline.

ii. Agents

Models that act rather than only answer: planning multi-step work, calling tools, and operating the same interfaces people use. Agents built on Gemini now browse, click and type their way through user interfaces with the Gemini 2.5 Computer Use model.

iii. Thinking

Giving a model room to reason before it responds: exploring several paths, discarding the weak ones, and committing to the strongest. Thinking is what turns a fluent model into a dependable one on hard problems in math, science and code.

II Google

Cameras that understand

Before DeepMind, I spent twelve years building AI products at Google. Two of them changed what a phone camera is for: searching what you see, and proving who you are.

i. Google Lens

I invented Google Lens. It began as an early prototype I showed to the VP of Engineering for Pixel phones at the Googleplex, and grew into the second most used search engine, behind only google.com.

The idea was simple: point your camera at something and ask about it. Search what you see.

ii. Pixel Face Unlock

I shipped Face Unlock on Pixel 4, then developed the industry's first Class 3 face biometric solution through a phone's selfie camera. Class 3 is Android's highest biometric tier: secure enough to approve mobile payments with a glance.

For four years, before moving to DeepMind, I owned all of Google's face biometrics.

iii. Synthetic identities

I pioneered the generation of synthetic human identities: faces of people who have never existed.

That technology became the cornerstone for avoiding identity drift in generative AI video, where a person has to remain the same person from the first frame to the last.

III On stage

Bringing agents to the enterprise

Frontier models matter when they reach the people who run real businesses. In 2026 I have been taking the agentic enterprise on the road, back home in Italy, with the Google Cloud team.

  1. Upcoming

    Codemotion Milan

    Presenting at the Tech Lead Summit, the invite-only track for tech leaders and CTOs, at Superstudio Più. See the sessions

  2. Upcoming

    Cloud AI Live Roma

    Speaking in two sessions at The Space Cinema Moderno: a day of demos, hands-on labs and production case studies on turning AI's potential into real solutions. If you will be in Rome, write to me and let's meet.

  3. Cloud AI Live Milano

    Presenting to executives focused on putting AI at the core of their enterprises: moving past isolated pilots toward agents that work alongside teams, end to end. Stop pilots, deliver value.

IV Personal projects

A builder's dream

It is an amazing time to be a product person and a thinker. If you have an idea, an AI agent can build it, and together you can refine it.

  • Research the economy of these EU regions and let's build an economic model.

    veneto2030.app

  • You want to have your own podcasts, but editing audio is painful.

    oneclickpod.app

  • You see a 3D model of Seoul and you wonder what Verona could look like.

    tiamoverona.vercel.app

V Origins

It started with silicon

Before models wrote code, I wrote the code that checked chips. At the Università degli Studi di Verona my research focused on functional test generation: automatically producing the inputs that prove a hardware design does what it should, before it is ever manufactured.

We built environments such as Amleto and Laerte++ for VHDL and SystemC designs, and asked whether genetic algorithms were the philosopher's stone of high-level test generation. The question I still work on is the same one: how do you make a machine's output trustworthy?

The path

From doctoral research in Verona to AI products in Mountain View.

  1. VeronaStudies and doctoral research in electronic design verification, Università degli Studi di Verona
  2. DresdenSemiconductor industry, Infineon Technologies
  3. AmsterdamEngineering and product leadership in EMEA
  4. Bay AreaEngineering and product leadership in Silicon Valley
  5. Mountain ViewAI product leadership at Google since 2013