DSC Europe 25 SCHEDULE

Fine-tuning pre-trained large language models (LLMs) offers a powerful and cost-efficient pathway to create high-performance, domain-specific Generative AI solutions. This session provides you with the practical knowledge and methodologies needed to harness fine-tuned models effectively, focusing on the complete lifecycle from data preparation to deployment. Participants will gain insights into how targeted, high-quality datasets can transform general-purpose models into specialized agents that excel in unique business and technical contexts. By the end of the session, attendees will understand both the strategic value and hands-on techniques required to translate fine-tuning into production-ready results.

In this talk, Dr. Arjun Jain will examine the evolution and integration of key subsystems essential to autonomous driving, including perception, localization, prediction, planning, and control. Dr. Jain will then dissect the critical technologies that enable vehicle autonomy and discuss their synthesis into cohesive systems capable of navigating complex environments. Special attention will be paid to the challenges of integrating these technologies, illustrated by the Boeing 737 redundancy issues, emphasizing the importance of robust, fail-safe mechanisms. Insights from Dr. Jain’s own research will be shared, focusing on practical applications and emerging solutions to address the ‘long tail’ problems of autonomous vehicle deployment. This talk will deepen the understanding of the current capabilities and future directions in autonomous vehicle technology.

Computer vision has become a cornerstone of innovation in the automotive industry, powering everything from driver-assist technologies to intelligent vehicle features. Combined with advances in machine learning, it enables cars to interpret their surroundings, make informed decisions, and deliver safer, smarter driving experiences. This talk will provide an overview of how computer vision is applied across the automotive domain today. The second part will focus on a real-world example: Rivian’s Gear Guard feature. This feature illustrates how end-to-end machine learning and computer vision pipelines are developed, iterated, and scaled for deployment in production vehicles. By walking through its evolution—from initial implementation to added functionalities and improvements—the talk will highlight the challenges and cycles involved in bringing a computer vision feature from concept to a robust, user-facing system.

In this talk, we will present our work on VerifAI system, an open-source biomedical question-answering system with unique mechanism to verify answers and detect hallucinations in generated answers. Namely, sciences, especially life sciences, have a low tolerance for non-factual information, and therefore many practitioners have been skeptical of using available tools, such as ChatGPT. While providing references to the information sources is a step in the right direction, it may not be enough, and even referenced generated answer may contain hallucinations. Therefore, we have developed a set of methods that on top of advanced RAG, combining lexical and semantic search capabilities and fine-tuning LLM with performance efficient fine-tuning method, such as LoRA, verify the answer and detect any remaining hallucinations. In order to increase efficiency, we have used quantization to improve latency and decrease hardware requirements for hosting the product. We will in detail discuss these techniques. The code, models and datasets generated during the project have been published in open-source and open science manner.