Sign In
Register

Partnership opportunities

Secure your seat

Call to action
Your text goes here. Insert your content, thoughts, or information in this space.
Button

Back to speakers

Pavithra
Rajendran
Senior Data Scientist, NLP & Computer Vision
Great Ormond Street Hospital
Pavi is currently leading the NLP and Computer Vision workstream at GOSH DRIVE. She has around 8+ years of academic and commercial experience as an NLP (Natural Language Processing) specialist. Previously, worked at a consulting firm and has experience using both traditional and deep learning-based NLP techniques for various client projects in both public and private sectors, from Proof-of-Concept to Production (Healthcare, Oil and Gas, Travel, Finance etc.) and, translating academic knowledge/research for commercial settings as well as latest commercial technologies by identifying the gaps and presenting with appropriate solutions for the business problems. She holds a Masters in Advanced Computing (Machine Learning,  Data Mining and High Performance Computing) and a PhD in Computer Science (Natural Language Processing).
Button
07 November 2024 14:30 - 15:00
Large language models in practice: Insights from building LLM-based pipelines for paediatric healthcare data
In this keynote, our speakers will provide an overview of the GOSH DRIVE team’s journey towards leveraging the capabilities of AI technologies to enrich our data resources. We will then focus on the potential capabilities of Large Language Models (LLMs) within the paediatric healthcare setting. Large volumes of data are recorded as part of routine healthcare, but often much of this data remains unused for the purposes of research and developing our capabilities as it is simply not feasible to manually extract all the data. In this session we will discuss how we have been able to perform high quality, reliable and accurate data extraction from documents, by leveraging the power of LLMs. We will share our experience and insights towards deploying "smaller" general-purpose LLMs for extracting relevant information that is present within unstructured texts and their enhanced zero-shot capabilities with the help of hand-crafted prompts. We will also share generalisability across a number of use-cases.