In this thought-provoking presentation, Ana will delve into the rapidly expanding realm of generative AI technology. With a focus on the latest advancements in text-to-image and large language models, Ana seeks to separate the hype from reality, and provide a comprehensive understanding of the most promising applications of this transformative technology. Additionally, the talk explores the significance of these developments for data scientists and technical professionals, providing valuable insights into the impact of generative AI on their roles.

Therese has been given access to the documentation of some of the Danish state's AI projects. The projects include how the state predicts long-term unemployment, and when there is a need to intervene in childcare cases. The common determinator is cases where ML has been used for ethically challenging domains. Therese gives her take on the quality and design of the public solutions.

Jonas works as a Data Scientist at Ekstra Bladet, focusing on Natural language processing and Machine Learning, and is a board member of Danish Data Science Community.

In his humorous talk, Jonas will share his best tips on what to do when presented with completely unreadable code and his take on how to practice Clean Code, as a Data scientist, in the best way possible.

Eleven months ago, I was assigned ownership of Scandinavia’s most complex card fraud detection model, and asked to overhaul it within 20 months.

The team I joined was under-staffed, working on unrelated deliveries, and had little to no documentation of the mode’s operation.

While the work isn’t complete yet, we are now in a significantly better place.

This talk will be about how we got there, by combining technical, business and process perspectives.

Visma’s ML Assets serve hundreds of thousands of customers serving over 1 million individual ML models in the process. Scaling the ML train-eval loop to this size has it’s own set of problems. In this talk we’ll dive into some of the techniques and concepts we use to manage ML quality for our model fleet.

Enterprises nowadays are utilizing billions of vision sensors to gather physical environment data, resulting in vast amounts of information. By combining this data with the power of AI, a wide range of problems can be solved in ways that were unimaginable a decade ago. However, modern vision AI networks are becoming more intricate to enhance accuracy and precision by encoding additional information, resulting in larger and more complex trained AI models with lower throughput and increased memory requirements. 

​

Machine vision applications face a new challenge with real-time AI inference due to this increase in size and complexity. NVIDIA has developed a two-fold solution to address this issue, which involves optimizing AI models for performance based on the hardware platform and maximizing the use of available accelerators.

This presentation explains how to utilize free software tools such as TAO Toolkit, DeepStream SDK, TensorRT, and Triton Inference Server to achieve real-time inference performance by leveraging NVIDIA's approach.

In recent years, generating content using artificial intelligence has become increasingly possible. This includes media such as sound, images (e.g., DALL-E 2), and most recently texts using ChatGPT. This keynote discusses the opportunities that come with these new technologies but also what negative effects they may have on society and our organizations. During the discussion, each technology will be demonstrated. We will look at how Trifork uses them in their daily operations to optimize processes and create new products for their clients.

The keynote is held by Mathias Grønne, a Machine Learning Specialist from Trifork. Besides working with Machine learning, Mathias is also part of the “Forum for Etisk AI” where the goal is to understand how technology may affect society and be ready for the upcoming AI regulation from the EU.

Unlock the secrets behind the next generation of AI agents, as large language models gain tool integration to boost their problem-solving skills and response quality.