The exhibition halls were a highlight of AMLD EPFL 2025, providing a space for valuable networking and knowledge exchange among attendees. Find out more about our exhibitors here buff.ly/y0d1kCv
We hope to see you at AMLD EPFL 2026!

Thank you to event partner Audemars Piguet for your support of AMLD EPFL 2025. We hope to see you again in 2026!

Thank you to event partner The AI Alliance for your support of AMLD EPFL 2025 and for the informative track “The Power of Collaboration: How the AI Alliance is Driving Open Innovation.”
We hope to see you again in 2026!

Thank you to event partner Giotto ai for your support of AMLD EPFL 2025 and to CTO Matteo Caorsi for your presentation at the track “AI and Cyber Defense: Strengthening National Security in the Digital Age.”
We hope you will join us again at AMLD EPFL 2026.

Thank you to event partner Google for your support of #AMLDEPFL2025, and to the team from Google for the engaging workshop “Building LLM Applications using Google Gemini - Long Context, Context Caching, and Retrieval Augmented Generation.”
We hope you will join us again at AMLD EPFL 2026.

Thank you to AMLD EPFL 2025 event partner Dataiku for your continued support of Applied Machine Learning Days. We hope to see you again at AMLD EPFL 2025!

Thank you to event partner Pictet for your support of AMLD EPFL 2025, and to the Pictet team for the informative track “Unlocking the Power of AI & GenAI in Financial Services: Strategies for Maximising Benefits.” We hope to see you next year at AMLD EPFL 2026!

Thank you to event partner Canton of Vaud for your support of #AMLDEPFL2025, and to Isabelle Moret for your informative keynote on the vibrant and innovative ecosystem of the canton. You can listen to the keynote here: youtu.be/8uoMrqB2Nd4?...
We look forward to seeing you at AMLD EPFL 2026!

Thank you to event partner Roche for your support of #AMLDEPFL2025. The Roche stand in the exhibition space was the place for many engaging and informative conversations with the Roche Diagnostics team. We hope you will join us again in 2026!

A big thank you to all of you for joining us at AMLD2025. Your contributions made this edition a great success. Experience the event again in our aftermovie! A special thanks to our partners, exhibitors, and speakers. See you next year for the 10th anniversary of AMLD, from February 10-13, 2026!

Thank you to Swisscom, a valued event partner for #AMLDEPFL2025, and Daniel Dobos, & the entire Swisscom team for delivering engaging and insightful track sessions. The Swisscom stand in the exhibition hall provided a meeting point for making connections & many meaningful conversations took place.

If you missed out on the keynote sessions at AMLD EPFL 2025, you can now listen to the speeches from Sepp Hochreiter, Nicolas Dessaigne, Mark Greaves, Nal Kalchbrenner, Florian Schweitzer, and Isabelle Moret on the AMLD YouTube channel. www.youtube.com/c/AppliedMac...

🎉 On Friday after 3 days of inspiring tracks, amazing keynotes, a hands-on workshop day, & an incredible lineup of speakers, startup pitches, poster exhibitions & exhibitor stands, we closed ​​ #AMLDEPFL2025. See you next year for AMLD's 10th anniversary, February 10-13, 2026!

🎉 On Friday after 3 days of inspiring tracks, amazing keynotes, a hands-on workshop day, & an incredible lineup of speakers, startup pitches, poster exhibitions & exhibitor stands, we closed ​​ #AMLDEPFL2025. See you next year for AMLD's 10th anniversary, February 10-13, 2026!

The answer? Leverage the knowledge built into VLMs (benefit from zero-shot and few-shot capabilities), use common off-the-shelf models and methods like fine-tuning and parameter-efficient tuning, and tap into multi-task capabilities

He then explores the shift from handwriting recognition to stroke interactions. While handwriting recognition is well-established, large models are driving a technological transformation. So how can we integrate handwriting, digital ink understanding & freeform capabilities into these models?

Claudiu Musat, researcher at Google DeepMind Switzerland, introduces strokes as a new modality for VLMs. Strokes enhance online expression and offer robustness against visual distractions. 🖌️

🌟 Kicking off the afternoon at #AMLD2025 with the "Natural Interactions with Foundation Models" workshop! Discover how to make AI feel as natural as real-world conversations—exploring text, speech, vision & strokes for intuitive, multimodal interactions. 💬✨ Let’s go beyond text! #LLMs

Now, our last stop before the lunch break: Create Artificial Worlds with PyTorch! 🌍✨

Join Vassilis Papadopoulos & Clement Honger as they explore Artificial Life models, from Cellular Automata to Neural Cellular Automata, using PyTorch & PyGame.

#AMLDEPFL2025 #PyTorch #MachineLearning

Gabriele Orlandi presented Bayesian regression and how it can optimize marketing techniques and contribute to better insights on marketing strategy effectiveness📈✨

When should you use context caching?
✅ Chatbots with extensive system instructions
✅ Repetitive analysis of lengthy video files
✅ Recurring queries over large document sets
✅ Frequent code repository analysis & bug fixing

Corpus-in-Context can be slow and potentially costly due to its quadratic complexity. This is where context caching comes to the rescue! When the corpus doesn't change much, computations can be performed once and reused multiple times. This significantly boosts speed and reduces costs. ⚡️

Now it's time to explore how to prototype LLM applications! The key principle: start simple and avoid over-engineering. Surprisingly, putting everything into context from the start works incredibly well. One of the strengths of Long Context LLMs is their ability to enable quick prototyping. But...

First, Max Buckley introduced the workshop and provided a quick refresher on key LLM concepts, such as tokenization, context windows, prefill or fine tuning. With the basics covered, the group is now ready to dive deep into the details and engage in hands-on practical activities.

💫 Let's join Google engineers for "Building LLM Applications using Google Gemini", a hands-on workshop on long context, context caching, and Retrieval-Augmented Generation (RAG)! Participants will learn how to prototype & productionize LLM apps effectively! 💻💡

For generating responses, we introduced chain-of-thought techniques like zero-shot prompting and in-context learning, which use new data during inference to improve output quality without retraining.

We highlighted efficient information retrieval for RAG systems, ensuring that relevant data is found for queries. Semantic search is vital for text embeddings and comparisons, while vector databases use kNN for quick retrieval and traditional filtering.🧑‍💻📝

Welcome to this last day of AMLD! ✨
We kick off with a Thomson Reuters workshop on how to apply generative AI in production with confidence📝🤖

We wrapped up the keynote with a lively panel discussion led by Marcel Salathé, featuring Nal Kalchbrenner and Florian Schweitzer. They explored how to find the next gems in the AI world.

Mackenzie Mathis presented Amadeus GPT, an AI breakthrough in behavioral analysis! 🗣️🧠 This AI system matches human-level performance using natural language and aims to mimic brain actions. 🤖 💫