About this Event
ATTENZIONE: Per rimanere sempre aggiornato su questo evento e sui prossimi, iscriviti alla nostra pagina Linkedin (https://www.linkedin.com/company/kaggle-practical-ai-milano).
L'evento si svolgerà in modalità ibrida: in presenza e online.
>>> IN PRESENZA
Luogo: Overnet Education Strada 3, Palazzo B4 Assago Milanofiori (MI)
Ingresso a partire dalle 18:15.
Evento gratuito con limite dei partecipanti e con registrazione obbligatoria.
>>> IMPORTANTE! ISCRIVITI A QUESTO EVENTO SOLO SE INTENDI PARTECIPARE IN PRESENZA. Se ti sei iscritto e per qualsiasi ragione sei impossibilitato a partecipare ANNULLA l'iscrizione.**
Se vuoi partecipare all'evento ONLINE segui le istruzioni riportate qui di seguito.
>>> ONLINE
ATTENZIONE LA DIRETTA STREAMING PER QUESTO EVENTO NON SARA' GARANTITA.
Iscriviti al nostro canale YouTube e segui l'evento in diretta streaming disponibile a questo indirizzo
https://www.youtube.com/@kagglePracticalAIMilano/streams
PROGRAMMA
18:15-18:30 Registrazione partecipanti
18:30-18:45 Presentazione della nostra community
18:45-19:45 Talk
19:45-20:30 Q&A e Networking
SPEAKER
Ivan Fioravanti, Founder, Investor, AI addicted Mensa member that has spent over 25 years in IT, from the data center to the cloud. He co-founded 4ward, a Microsoft System Integrator, in 2001, where he was CTO. In 2014 he co-founded CoreView, a Microsoft 365 management and security platform, leading technology and product for twelve years. He is still a shareholder and board member there, but in 2026 he stepped back from operations to deep dive full time on Local AI.He tried to run open models locally since the first Mac Studio M2 Ultra, and has worked closely with Apple's MLX team and contributed to the framework. He shares benchmarks, experiments, collaborate with various OSS AI related projects and share everything on LinkedIn and X (@ivanfioravanti), where he is one of the most vocal voices in the Local AI community. He has run models of up to 750 billion parameters on hardware you can keep on a desk.
ABSTRACT
Most people think of generative AI as a cloud API, token based solution, that’s it. This talk shows the other side: AI stack that runs entirely on hardware you can own, with zero calls to cloud AI services. All live demos (let’s break things!)! Very few slides.On stage we'll put together a small "mini data center": two Mac Studio M3 Ultra with 512 GB of unified memory each, a MacBook Pro M5 Max, a cluster of four NVIDIA DGX Spark and a LuceBox. We'll use them to generate every kind of content, live: text and code with open models above 300 billion parameters, images, speech, music and video. Each demo runs across several machines. For example, a large language model on one machine writes the prompts and text for a newspaper, and another machine turns them into images and podcast in seconds.We’ll see how to use herdr with pi to push hardware and models to the max and see it in action.Beyond the demos, the goal is having an open talk to. cover what we need to know before entering the Local AI arena:
- what really fits in memory
- quantization trade-offs
- splitting very large models across multiple machines
- Apple Silicon vs. NVIDIA, and when each one makes sense
- where local models still lose to frontier models in the cloud
- whatever else came int out minds during the session
We'll also cover the business case. Many workloads don't need a real-time answer: code and pull request reviews, bug triage, call and document analysis, drafting, and image generation can all run as overnight batch jobs. Moving them to local hardware can cut more than $100,000 a year in cloud costs.We you'll leave with a clear picture of what Local AI can do today and a practical way to adopt it: local by default, cloud when it really matters.
Event venue & nearby stays
OverNet, Strada 3, Assago, Italy
Concerts, fests, parties, meetups — all the happenings, one place.