Speaking
Things I like to talk about
Upcoming appearances, talks I've already given, and topics I'd be happy to bring to a future event.
Upcoming
Beyond Benchmarks: What ARC Teaches Us About Measuring Intelligence
Codemotion Milan 2026
Most AI benchmarks measure performance on tasks we can easily collect, score and scale. But do they measure intelligence? Starting from François Chollet’s 2019 paper On the Measure of Intelligence, this talk revisits a century-old question through a modern engineering lens: what should it mean for a machine to learn efficiently, generalize from few examples and solve genuinely novel problems? We will trace the path from historical ideas of intelligence measurement to ARC, the Abstraction and Reasoning Corpus: a benchmark designed around abstract reasoning, minimal priors and tasks that are easy for humans but hard for current AI systems. The goal is not to debate AGI hype, but to understand why benchmark design shapes research, product expectations and our mental models of progress. Attendees will leave with a clearer framework for reading AI benchmarks critically and asking better questions about reasoning claims.
Past talks
Estrarre informazioni strutturate dalle immagini con Pydantic AI e FastAPI
DevFest Roma 2025
A talk about how the Computer Vision Engineer role is changing with vision-enabled models and multimodal LLMs. It shows how to build a small Pydantic AI and FastAPI service that receives an image, queries a generative model, and returns validated, typed data.
Flying SOLID-ly: Python design principles explained with birds
PyData Roma Capitale x R-Ladies Rome x Rome R Users Group meetup @ INGV
A practical talk about SOLID principles in Python, using birds as a playful way to make software design concepts easier to reason about.
Computer Vision IRL: From Idea to On-Premise Deployment
PyCon Italia 2024
A real-world case study from Immobiliare.it on turning a business need into a deployed image classification service for real estate. The talk covered requirements, dataset creation, model development and optimization, deployment with FastAPI and TorchServe, and production monitoring.
Still on my mind
Topics I keep thinking about, testing, or trying to explain better.
AI tools for coding
How I use AI tools to write, review, debug, and ship code faster without outsourcing judgment.
What agents actually are
Introductory talks that cut through the hype: tools, loops, memory, planning, and where agents are useful in real products.
What we are learning while building AI features for game-based learning: product constraints, user value, and boring production details.
Community
I help organize PyData Roma Capitale and enjoy formats where engineering people can share practical work without over-polishing it. Workshops, lightning talks, and cross-community meetups are usually the most fun.