Session Previews 2024
ML Week Europe 2024: Unveiling the Director's Favourite Sessions
Martin Szugat
Discover the personal favorite sessions and keynotes of Martin Szugat, the esteemed Program Director of Machine Learning Week Europe!
Prepare to be captivated as he presents his selection of must-attend sessions. Dive into these remarkable talks to uncover the reasons behind his enthusiasm and find out why you simply can’t afford to miss them!
Keynote: xLSTM: New Architectures for Large Language Models
Maximilian Beck
Today’s LLMs such as ChatGPT show an impressive performance and have the potential to revolutionize our daily life. All these LLMs are based on the Transformer architecture with the Attention mechanism at its core. Due to the quadratic scaling with context length, Attention makes processing of long sequences very expensive. In this talk Maximilian presents xLSTM, a novel architecture for LLMs that scales only linear in context length while still outperforming Transformers on language modeling.
Duplicate Record Detection Using GenAI Techniques to Improve Data Quality
Ian Ormesher
Duplicate records can have a negative impact on many areas of a business. Current methods to detect duplicate records use traditional NLP techniques known as “Entity Matching”. An improvement to this traditional method can be achieved by incorporating GenAI techniques that do not entail any calls to OpenAI. Not only does this produce better matches, but it also keeps the data safe, since no information is transferred externally.
Developing an LLM-Powered Conversational Agent for External Users at Feedly: Challenges and Insights
Farah Ayadi
In this session, we will share our experience developing a context-aware conversational agent using Large Language Models (LLMs) at Feedly AI search platform. The agent offers intelligent services like summarization, analysis, and report generation, tailored to the user’s context and content. We will discuss the unique challenges and strategies involved in catering to external users with stringent requirements, where accuracy, consistency, and the ability to synthesize large amounts of data are crucial, unlike most current initiatives focusing on internal enterprise use cases with higher tolerance for inconsistency and simpler information retrieval.
Realtime Anomaly Detection for Better Decision-Making at Angelini Technologies - Fameccanica
Enrico Iavazzo
Managing production facilities with over 5000 variables can be a challenging task. Enrico will present a system that enables real-time data analysis, displaying only the anomalies that require the attention of domain experts on the HMI. Output is available for the PLC, enabling closed-loop applications. This approach enables a better decision-making ensuring that any potential issues are resolved before they occur, thereby saving time and resources.
How To Open Ears, Minds, Doors And Budgets For Data Products With Data Storytelling – Learnings from a Pharma Company
Jack Lampka & Dr. Julia Zukrigl
Of all the data projects that fail, 80% fail due to poor communication. When decision makers don’t understand a data product, they don’t need more information: they need less of it. In this talk, Jack and Julia will show how data storytelling is helping to promote data products and get buy-in at a pharma company. Tune in to learn the process of shaping the narrative and finding a data product marketing strategy. Get expert tips on how you can use data storytelling for your own data / AI product.
Human-in-the-loop: Practical Lessons for Building Comprehensive AI Systems
Miha Garafolj
Machine learning models should not live in a vacuum. When a predictive model is exposed to the user to enrich a scientific or industrial workflow, incorporating human feedback is essential for the model to improve over time, overcome distribution shifts, and learn novel phenomena. We address the engineering challenges of building interactive active learning systems on a practical example of large-scale video analysis.
Optimizing LLM for Endress+Hauser: An Innovative Approach to Chatbot Design with Azure Semantic & Vector Search
Muhammad Saad Uddin
In this session, we will have an in-depth exploration of chatbot innovation applied in a real-world setting at Endress+Hauser. This session will focus on our practical implementation of Large Language Models, combined with the capabilities of Azure’s Semantic & Vector Search and strategically advance Retrieval-Augmented Generation (RAG) approach, to build a chatbot that is not only highly responsive but also remarkably intuitive .We will provide a comprehensive overview of our approach, detailing the process of integrating technology stacks and navigating the complexities of chatbot design, demonstrating the tangible benefits of these technologies in a real-world, industrial setting.
How AMP Managed Change: Step by Step from Paper-Free Towards Predictive Maintenance
Wiebke Apitzsch & Andreas Engel
70-80% of all digital transformations fail, we know this for decades. While IT is often blamed for it, tech is unlikely to be the root cause: Most transformations fail due to people issues. When setting up AMPs digital transformation in production, we closely connected the tech agenda with change. We will share how: the plant team was involved, we kept leadership involved and data helped to gain insights. Paper-free builds the data baseline, it serves as the main example.
How MLOps Practices Speed Up the Development of Condition Monitoring Services at KSB SE & Co. KgaA
Simon Kneller
At KSB SE & Co. KGaA, the number of productive ML-models increased with the growing customer base of its digital product KSB Guard. As a result, the need for a clear separation of responsibilities between data scientists and operations increased. In particular, the further development of already productive models became a pain point. In this talk I want to present a service architecture that promotes seamless collaboration between data science and operations teams.
Guarding Public Works Projects with an Early Warning System for the Hong Kong Development Bureau
Dr. Michele Dallachiesa
This talk discusses the Government of Hong Kong’s challenges in overseeing and assuring its 500+ active capital projects, aiming to avert costly delays and budget overruns with a suite of ML models. We introduce the development of an early warning system, drawing insights from 2,700 years of cumulative construction activity and a total value of USD 60bn. We will present and discuss our problem-solving approach, experimental evaluation, and conclude with useful guidelines for forecasters.
Lean Experimentation and Decision Science for Warehouses at Delivery Hero
Sayon Kumar Saha
An effective picker schedule in warehouses in Quick-Commerce addresses the challenge of resource utilization while maintaining low delivery time. The talk focuses on the challenges in running decision science experiments with humans involved, and embracing learnings by strategizing: How to utilize limited and unclean data caused by stochastic activities and human-error in warehouses? How to draw insights while navigating uncertainties? How to optimize operational decision making processes?
Interview
Rohit Agarwal
Q: What makes your job so exciting?
Working as a Chief Data Officer at Bizom, a top SaaS company in the Indian retail sector, is truly thrilling. Our access to vast amounts of retail data enables us to uncover fascinating insights into consumer behavior in almost real-time. Additionally, my team constantly explores cutting-edge technologies and their potential applications in the retail industry, making our work a blend of research and practical implementation.
Q: What are the critical challenges in data science?
The primary challenge in data science lies in data quality. We frequently encounter duplicate, incomplete, and erroneous data, making it difficult to derive accurate insights and introducing potential risks. Ensuring that the data is cleaned up and analyzed properly is crucial, especially since our reports are utilized by financial institutions.Furthermore, the rapid pace of change in the field is staggering. In just 1.5 years, generative AI has become a dominant force, overshadowing core machine learning algorithms like Random Forest that were once significant in discussions.
Q: What is the most promising use case of Generative AI for your industry / business?
One significant application of Generative AI in our industry is the generation of product descriptions for listing on e-commerce platforms, where there is often incomplete information available. By using product images as input prompts, AI can assist in automatically creating detailed and accurate product descriptions, saving time and effort.
Another valuable use case involves report summarization, allowing users to input queries in natural language, such as “What is the year-over-year sales comparison between outlet #1 and outlet #2?” Extracting this specific information from lengthy reports can be a time-consuming task. Implementing Generative AI algorithms like text-to-SQL can streamline this process by quickly generating answers to such specific queries.
Q: Sneak preview: Please tell us one take-away that you will provide during your talk at Machine Learning Week Europe
During my session at Machine Learning Week Europe, I will share insights on creating a foundational architecture utilizing Vector Database & Embeddings to enhance intelligent search capabilities. This includes discussing various options for text, image, and multi-modal searches. Additionally, I will showcase real-world applications that harness this architecture and discuss potential directions for further research and development.
Interview
Prince Tyagi
Q: What makes your job so exciting?
A: Being an AI engineer at a startup is thrilling due to the ever-changing nature of each day. I have the opportunity to create effective solutions that enhance our product, analyze complex data, and implement new algorithms to elevate our product in practical scenarios.
Q: What are the critical challenges in data science?
A: The most challenging aspects of data science include ensuring data quality and integrity, developing algorithms/models that are easy to interpret, and integrating new algorithms without causing any disruptions to current workflows.
Q: What is the most promising use case of Generative AI for the healthcare industry?
A: Through the use of Generative AI(RLHF), it is possible to predict potential health risks caused by limited movement by analyzing tracked ergonomic data. This can be accomplished by improving generative AI models based on feedback received from healthcare experts and individuals regarding their posture and ergonomic practices. The proactive approach is designed to improve patient care by addressing the risks associated with a sedentary lifestyle and offering personalized interventions to enhance health outcomes.
Q: Sneak preview: Please tell us one takeaway that you will provide during your talk at
Machine Learning Week Europe.
A: Come join me at Machine Learning Week Europe as I discuss how RLHF techniques can align LLMs to create high-quality outputs that effectively meet human expectations. This cutting-edge approach not only improves model performance but also enhances their adaptability and responsiveness to real-world situations.
Interview
Dr. Monika Lieb
Q: What makes your job so exciting?
A: As a Data Scientist at NIQ, the world’s largest consumer intelligence provider, I have the privilege of accessing one of the most valuable datasets globally. Wearing multiple hats, I combine expertise in statistics, programming, and domain knowledge to translate business questions into meaningful analyses. The excitement lies in uncovering valuable insights, interpreting market dynamics, and predicting behaviour, thereby ultimately shaping our clients’ strategic decisions.
Q: What are the critical challenges in data science?
A: Every valuable insight a data scientist provides relies on high-quality data. While improving data collection processes from diverse sources and managing data complexity may not be the most glamorous aspects of data science, they are becoming increasingly vital. Additionally, it’s crucial to translate these insights into business language for non-data scientists, especially as the methods applied grow increasingly complex. This translation is key to ensuring that data science can provide meaningful value.
Q: What is the most promising use case of Generative AI for your industry / business?
A: In the market research domain, integrating human intelligence and comprehensive, granular, up-to-date data with Generative AI can lead to significant enhancements across various areas. Use cases include advanced personalization tailored to our client needs, acceleration of product innovation cycles via empowering human creativity, or data quality assurance via AI-driven automation of data discrepancy detection and correction.
Q: Sneak preview: Please tell us one take-away that you will provide during your talk at Machine Learning Week Europe.
A: At Machine Learning Week Europe this November you will learn from me how our team at NIQ has accomplished to model online marketplace sales. You will discover the importance of identifying all available and relevant information, integrating diverse datasets and leveraging information synergies.