Valeria Zuccoli
Data Scientist
Milan, Italy
Actions
Statistician by education, Valeria is a Data Scientist specializing in real-time models for complex, challenging scenarios. Driven by a deep curiosity for the latest research, she applies advanced analytical techniques to build intelligent systems that deliver immediate and effective solutions in high-impact domains.
Links
Area of Expertise
Topics
The Zero-Temperature Lie: Why Your Deterministic LLM is Still Hallucinating Randomness
Deterministic LLM inference sounds simple: set temperature to zero and get consistent answers. But reality is messier, and asking the same question a thousand times might yield dozens of different responses.
While GPU parallelism and floating-point arithmetic play a role, a study by Thinking Machine Labs identifies the real culprit as… other users. Concurrent requests to the same server change the batch size, which silently alters the order of transformer computations, making the output look like uncontrollable randomness.
This talk will debunk the common myths about LLM nondeterminism, walk through the actual mechanics of inference engines, and explain what it would take to achieve true reproducibility.
Whether calling an external API or running a self-hosted inference stack, attendees will leave with a clear understanding of why this happens, which strategies can help address it, and how to think about reproducibility in AI systems.
Delivered at PyConIT 2026
Angels & Demons: A Fact-Based Look at the Consequences of AI
The conversation about AI’s impact is often driven by hot takes, fear, or hype.
This talk is an evidence-first tour of what we know so far, using peer-reviewed research and reliable reports to test the claims that fuel both optimism and pessimism.
We’ll examine four areas:
- Work & Economics — Are we seeing AI-driven job losses? What do hiring signals say about junior roles?
- Software Development — What do controlled studies show about AI-assisted coding: productivity, quality, and security trade-offs?
- Education — How widespread is AI-enabled cheating, and how does that compare with evidence from pilots where AI is used as a tutor?
- Human Relationships — Where do chatbots help, and where do safety risks emerge in high-stakes situations?
The audience will leave with a clear mental model, a set of concrete references, and a grounded view of the real consequences of AI — including what’s proven, what’s uncertain, and what’s still speculation.
Delivered at Codemotion Roma 2026
The Dark Side of GenAI: Unmasking AI Role in the Fake News Epidemic
Artificial Intelligence is no longer just a tool for analysis; it's now a powerful generator, capable of crafting realistic content at an unprecedented scale. This session explores the concerning evolution of AI's role in creating and spreading fake news.
We'll begin by looking at the power of AI-driven content generation and its implications for public discourse. Then, we'll move from theory to practice, demonstrating how to vibe code a convincing fake newspaper website in a matter of minutes.
By the end of this talk, attendees will gain a critical understanding of the urgent challenges and the defenses we need to navigate our evolving information landscape.
Delivered at DataBeers Padova #12
Repetita Non Iuvant: Why Generative AI Models Cannot Feed Themselves
As AI floods the digital landscape with content, what happens when it starts repeating itself?
This talk explores model collapse, a progressive erosion where LLMs and image generators loop on their own results, hindering the creation of novel output.
We will show how self-training leads to bias and loss of diversity, examine the causes of this degradation, and quantify its impact on model creativity.
Finally, we will also present concrete strategies to safeguard the future of generative AI, emphasizing the critical need to preserve innovation and originality.
By the end of this talk, attendees will gain insights into the practical implications of model collapse, understanding its impact on content diversity and the long-term viability of AI.
First delivered at Py4AI 2025, Pavia, Italy
Also held at AI Heroes 2025, Turin, Italy
Also held at AI Conf 2025, Milan, Italy
Also held at PyData Paris 2025, Paris, France
Also held at Codemotion Milan 2025, Milan, Italy
Codemotion Roma 2026 Sessionize Event
Databeers Padova #12
Databeers Padova è il chapter patavino di Databeers, una comunità internazionale di appassionati del mondo dati.
Databeers organizza eventi pubblici, informali e aperti a tutti per divulgare la cultura del dato in maniera aperta e accessibile: parliamo di dati bevendoci una buona birra!
Codemotion Milan 2025 Sessionize Event
PyData Paris
PyData brings together data scientists, engineers, educators, developers, and maintainers of opens-source projects, for in-depth training, insightful keynotes, networking events, and practical talks.
AI Conf
Starting in 2019, AI Conf is the first conference designed to bring Artificial Intelligence topics to Italy.
AI Heroes
AI Heroes is the Italian Artificial Intelligence Conference for Developers, Engineers, Data Scientists & Product Managers.
Py4AI
Py4AI is an international conference where Python and AI enthusiasts connect, learn, and explore the latest advancements in the field.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top