2024-06-11 –, Kesselhaus
GenAI is transforming organizations rapidly. But how can we scale individual experimentation to enterprise-grade business products? This talk will cover learnings from hundreds of use cases in pharma about user intentions, information access patterns and quality metrics.
GenAI is the latest buzzword promising to bring transformative capabilities to enterprise users at ready stage. With the current wave of commodity GenAI models, such as ChatGPT, organizations are tempted to translate all their business needs into prompts and deploy them as conversational chatbots. However, there are manifold challenges blocking us to unlock the full potential of GenAI, especially from the perspective of expert users with knowledge-based goals.
This talk will cover learnings and techniques our pharma organization has developed from translating hundreds of use cases into GenAI prototypes. We introduce six common groups of user goals that we have identified in a study of 85 proposals and evaluated in dozens of production use cases. We break down their most critical challenges and propose reusable architecture patterns for speeding up prototyping, accessing corporate information and measuring result quality.
Sebastian Arnold graduated in Computer Science at TU Berlin and received his Ph.D. in 2020 for his thesis on Machine Reading for Domain-specific Text Resources from University of Fribourg, Switzerland in cooporation with the DATEXIS research group at Beuth University of Applied Sciences Berlin. From 2020 he led the data product development of Curalie GmbH (Fresenius Group) in the field of digital healthcare as Head of Data Science. In 2023 he joined Bayer AG Pharmaceuticals as Principal Data Scientist and co-leads the organization's cross-divisional platform for Generative AI.