Search – talk-data.com

Title & Speakers	Event
PyData Rhein-Main I AI Agent Hacks & GTC 2025 Insights 2025-03-27 · 16:30 Topic: AI & Data Science in practice Venue: In person in Darmstadt and live on PyData.TV on YouTube Agenda 5:30 pm Doors open 6:00 pm Welcome 6:15 pm How to Hack an Agent – or Not · Thomas Fraunholz 6:45 pm Networking with snacks and beverages 7:45 pm AI & Data Strategy after NVIDIA GTC 2025 · Alexander C. S. Hendorf 8:15 pm Lightning Talks 8:30 pm Networking with snacks and beverages 9:00 pm End 🍿 How to join remotely Talk#1 - Thomas: https://youtube.com/live/pTSKL6e66mE Talk#2 - Alexander: https://youtube.com/live/ooyT412QCSI ⚡️ Lightning Talks Feel free to submit a proposal How to sign up for on site It's important for us to make this meet up happen in a responsible way. We have limited seats available only. No limits to sign up remotely! This event will be in English. ---- Talk #1 How to Hack an Agent – or Not Thomas Fraunholz Senior Researcher AI @ Smart Labs AI Large language models (LLMs) are not as secure as they seem. Beyond their tendency to “hallucinate,” they can be manipulated using jailbreaks and adversarial prompts, bypassing safeguards designed to keep them in check. But the real challenge arises when LLMs are connected to agents with real-world capabilities—like sending emails. This talk explores the security risks of AI agents and the ongoing research into making them more resilient. Using the "Adaptive Prompt Injection: LLMail Inject" challenge from the IEEE Conference on Secure and Trustworthy Machine Learning as a case study, we’ll examine how Microsoft’s Phi3 and OpenAI’s GPT-4o-mini handle adversarial attacks. We’ll break down security techniques like LLM judges, task drift detection, and prompt shields—critical concepts as the EU AI Act's security mandates take effect in August 2025. Attendees will gain insights into the strengths and weaknesses of current AI security mechanisms and learn practical strategies for assessing the safety of AI agents in production environments. About the Speaker Thomas is an MLOps and NLP expert with a background in applied mathematics and embedded programming. He has led two publicly funded AI research programs with the German Aerospace Center and is currently focused on AI-driven cybersecurity at Smart Labs AI GmbH. In his spare time, he’s developing a low-budget drone system for detecting bark beetle infestations, blending his passion for AI and embedded systems. Talk #2 AI & Data Strategy After NVIDIA GTC 2025: What You Need to Know Alexander C. S. Hendorf AI & Data Strategy and Implementation @ opotoc GmbH NVIDIA GTC is one of the most influential conferences in AI, showcasing advancements in accelerated computing, robotics, healthcare, and finance. In this talk, Alexander Hendorf will share his key takeaways from the conference, focusing on what’s most relevant for the community. From cutting-edge AI hardware and model optimization to real-world applications in robotics, healthcare, and financial modeling, we’ll explore how these technologies impact data workflows. Expect insights on practical AI adoption, the latest breakthroughs in GPU acceleration, and what’s actually useful (vs. just hype) for data scientists and engineers. About the Speaker Alexander is a data intelligence and AI expert with over 20 years of experience in digitalization and data-driven decision-making. As an independent consultant, he's specialized in AI & data strategy and implementation. A frequent speaker and chair at international conferences like PyCon DE, PyData Berlin, and EuroPython, he is also a Python Software Foundation Fellow and EuroPython Fellow. He serves on the board of the Python Software Verband and, since 2024, has also been leading Pioneers Hub, a non-profit dedicated to supporting tech communities. ---- Acknowledgements Also a big thank you to our partners: Hessian.AI, for hosting the meetup. PIONEERS HUB, for organising. Contact If you have any questions or suggestions, please feel free to contact us via: Meetup Want to speak? Submit a talk here. Interested in hosting an event? Here's our Info-Deck & contact to the organisers!	PyData Rhein-Main I AI Agent Hacks & GTC 2025 Insights
PyData Southampton - 14th Meetup 2025-03-18 · 19:00 Venue: Carnival House, 100 Harbour Parade, Southampton, SO15 1ST 📢 Want to speak 📢: submit your talk proposal Main Talks 1️⃣ How to Hack an Agent - or not, Thomas Fraunholz LLMs are not safe. And it's not because of their ability to hallucinate while working. LLMs can be manipulated. So far, the known security mechanisms have not really proven to be solid. For example, if you use jailbreaks correctly, you can get an LLM to do things it is not supposed to do. But hey, what's stopping us from giving an LLM a little email summary? Nothing. Unless you connect it to an agent that can also send emails. Which brings us to the issue: Can an AI agent be secured where even an LLM cannot be secured? We want to approach this question by looking at state-of-the-art security concepts, from catching LLM task drift with activations, to prompt shields, spotlighting, and command hierarchy. Is my AI agent safe? Let's find out! 2️⃣ DeepICD: Deep learning for assessing one’s eligibility for an implantable cardioverter-defibrillator - Samuel Ward This talk will begin by laying out the problems of surgical decision making and inappropriate shocks involved with implantable cardioverter-defibrillators. We will then present our deep learning tool and discuss how it may begin to solve these problems. Along the way we will discuss neural network architecture design, the training process and how to appropriately evaluate the model. Lightning Talks ⚡ ⚡1️⃣ TBD ⚡2️⃣ TBD Please note: 🚨🚨🚨A valid photo ID is required by building security. You MUST use your initial/first name and surname on your meetup profile, otherwise, you will NOT make it on the guest list! 🚨🚨🚨 This event follows the NumFOCUS Code of Conduct, please familiarise yourself with it before the event. If your RSVP status says "You're going" you will be able to get in. No further confirmation required. You will NOT need to show your RSVP confirmation when signing in. If you can no longer make it, please unRSVP as soon as you know so we can assign your place to someone on the waiting list. * Code of Conduct: This event follows the NumFOCUS Code of Conduct, please familiarise yourself with it before the event. Please get in touch with the organisers with any questions or concerns regarding the Code of Conduct. * There will be pizza & drinks, generously provided by our host, Carnival UK. * Logistics** Doors open at 6.30 pm, talks start at 7 pm. For those who wish to continue networking and chatting we will move to a nearby pub/bar for drinks from 9 pm. Please unRSVP in good time if you realise you can't make it. We're limited by building security on the number of attendees, so please free up your place for your fellow community members! Follow @pydatasoton (https://twitter.com/pydatasoton) for updates and early announcements. We are also on Instagram/Threads as @pydatasoton, and find us on LinkedIn.	PyData Southampton - 14th Meetup

Topic: AI & Data Science in practice Venue: In person in Darmstadt and live on PyData.TV on YouTube

Agenda 5:30 pm Doors open 6:00 pm Welcome 6:15 pm How to Hack an Agent – or Not · Thomas Fraunholz 6:45 pm Networking with snacks and beverages 7:45 pm AI & Data Strategy after NVIDIA GTC 2025 · Alexander C. S. Hendorf 8:15 pm Lightning Talks 8:30 pm Networking with snacks and beverages 9:00 pm End

🍿 How to join remotely Talk#1 - Thomas: https://youtube.com/live/pTSKL6e66mE Talk#2 - Alexander: https://youtube.com/live/ooyT412QCSI

⚡️ Lightning Talks Feel free to submit a proposal

How to sign up for on site It's important for us to make this meet up happen in a responsible way. We have limited seats available only. No limits to sign up remotely!

This event will be in English.

----

Talk #1 How to Hack an Agent – or Not Thomas Fraunholz Senior Researcher AI @ Smart Labs AI Large language models (LLMs) are not as secure as they seem. Beyond their tendency to “hallucinate,” they can be manipulated using jailbreaks and adversarial prompts, bypassing safeguards designed to keep them in check. But the real challenge arises when LLMs are connected to agents with real-world capabilities—like sending emails. This talk explores the security risks of AI agents and the ongoing research into making them more resilient. Using the "Adaptive Prompt Injection: LLMail Inject" challenge from the IEEE Conference on Secure and Trustworthy Machine Learning as a case study, we’ll examine how Microsoft’s Phi3 and OpenAI’s GPT-4o-mini handle adversarial attacks. We’ll break down security techniques like LLM judges, task drift detection, and prompt shields—critical concepts as the EU AI Act's security mandates take effect in August 2025. Attendees will gain insights into the strengths and weaknesses of current AI security mechanisms and learn practical strategies for assessing the safety of AI agents in production environments.

About the Speaker Thomas is an MLOps and NLP expert with a background in applied mathematics and embedded programming. He has led two publicly funded AI research programs with the German Aerospace Center and is currently focused on AI-driven cybersecurity at Smart Labs AI GmbH. In his spare time, he’s developing a low-budget drone system for detecting bark beetle infestations, blending his passion for AI and embedded systems.

Talk #2 AI & Data Strategy After NVIDIA GTC 2025: What You Need to Know Alexander C. S. Hendorf AI & Data Strategy and Implementation @ opotoc GmbH NVIDIA GTC is one of the most influential conferences in AI, showcasing advancements in accelerated computing, robotics, healthcare, and finance. In this talk, Alexander Hendorf will share his key takeaways from the conference, focusing on what’s most relevant for the community. From cutting-edge AI hardware and model optimization to real-world applications in robotics, healthcare, and financial modeling, we’ll explore how these technologies impact data workflows. Expect insights on practical AI adoption, the latest breakthroughs in GPU acceleration, and what’s actually useful (vs. just hype) for data scientists and engineers.

About the Speaker Alexander is a data intelligence and AI expert with over 20 years of experience in digitalization and data-driven decision-making. As an independent consultant, he's specialized in AI & data strategy and implementation. A frequent speaker and chair at international conferences like PyCon DE, PyData Berlin, and EuroPython, he is also a Python Software Foundation Fellow and EuroPython Fellow. He serves on the board of the Python Software Verband and, since 2024, has also been leading Pioneers Hub, a non-profit dedicated to supporting tech communities.

---- Acknowledgements Also a big thank you to our partners:

Hessian.AI, for hosting the meetup.
PIONEERS HUB, for organising.

Contact If you have any questions or suggestions, please feel free to contact us via:

Meetup
Want to speak? Submit a talk here.
Interested in hosting an event? Here's our Info-Deck & contact to the organisers!

Venue: Carnival House, 100 Harbour Parade, Southampton, SO15 1ST 📢 Want to speak 📢: submit your talk proposal

Main Talks 1️⃣ How to Hack an Agent - or not, Thomas Fraunholz LLMs are not safe. And it's not because of their ability to hallucinate while working. LLMs can be manipulated. So far, the known security mechanisms have not really proven to be solid. For example, if you use jailbreaks correctly, you can get an LLM to do things it is not supposed to do. But hey, what's stopping us from giving an LLM a little email summary? Nothing. Unless you connect it to an agent that can also send emails. Which brings us to the issue: Can an AI agent be secured where even an LLM cannot be secured? We want to approach this question by looking at state-of-the-art security concepts, from catching LLM task drift with activations, to prompt shields, spotlighting, and command hierarchy. Is my AI agent safe? Let's find out!

2️⃣ DeepICD: Deep learning for assessing one’s eligibility for an implantable cardioverter-defibrillator - Samuel Ward This talk will begin by laying out the problems of surgical decision making and inappropriate shocks involved with implantable cardioverter-defibrillators. We will then present our deep learning tool and discuss how it may begin to solve these problems. Along the way we will discuss neural network architecture design, the training process and how to appropriately evaluate the model.

Lightning Talks ⚡ ⚡1️⃣ TBD ⚡2️⃣ TBD

Please note:

🚨🚨🚨A valid photo ID is required by building security. You MUST use your initial/first name and surname on your meetup profile, otherwise, you will NOT make it on the guest list! 🚨🚨🚨
This event follows the NumFOCUS Code of Conduct, please familiarise yourself with it before the event.

If your RSVP status says "You're going" you will be able to get in. No further confirmation required. You will NOT need to show your RSVP confirmation when signing in. If you can no longer make it, please unRSVP as soon as you know so we can assign your place to someone on the waiting list.

*** Code of Conduct: This event follows the NumFOCUS Code of Conduct, please familiarise yourself with it before the event. Please get in touch with the organisers with any questions or concerns regarding the Code of Conduct. *** There will be pizza & drinks, generously provided by our host, Carnival UK. ***

Logistics Doors open at 6.30 pm, talks start at 7 pm. For those who wish to continue networking and chatting we will move to a nearby pub/bar for drinks from 9 pm.

Please unRSVP in good time if you realise you can't make it. We're limited by building security on the number of attendees, so please free up your place for your fellow community members!

Follow @pydatasoton (https://twitter.com/pydatasoton) for updates and early announcements. We are also on Instagram/Threads as @pydatasoton, and find us on LinkedIn.

talk-data.com

People (168 results)

Activities & events