ATS Seminar Series: Martin Foltin and Rangan Sukumar
The Advanced Technologies Section (ATS) of the National Center for Computational Sciences at ORNL is a world leader in developing and deploying scientific and technical solutions for leadership-class computing environments. The R&D activities of ATS are organized around designing and deploying leadership class systems, developing artificial intelligence solutions for science and smart facilities for the future, and stewardship of data and workflows at scale to enable science. The ATS Seminar Series is a forum for learning from experts and engaging with collaborators to advance their scientific mission.
Enabling Search and Discovery with AI for Science
From being logic-hungry in the Big Data era to the golden age of success stories with (data-hungry) deep learning, we are amidst the metamorphosis of data-centric artificial intelligence (AI) to a knowledge-nurtured future. Acknowledging the potential of scientific data as drivers of economies in that future (in use-cases such as precision medicine, self-driving cars, supply chain logistics, etc.) and AI as the ‘key’ to unlocking timely insights in the Exascale era, this talk posits the following: (i) scientific discovery requires multiple modes of search (i.e. beyond O(N) traversal algorithms to higher order transformations and pattern recognition, multi-modal relationship extraction, etc.), (ii) state-of-the-practice databases and tools are expensive, data lakes are slow and MLOps platforms can be unwieldy for different shapes of scientific data (images, time-series, sequences, N-d dense and sparse, 3D points and meshes), (iii) workflow efficiency (that includes ease of query expression, orchestration, experimentation and reproduction) is as important as raw hardware and software performance, and (iv) computational scientists who thrive on interpretability, explainability and reproducibility of results can significantly benefit from a platform that automatically manages metadata and lineage in complex end-to-end AI pipelines.
As an invitation for open collaboration addressing such grand challenges around data in the Exascale era, this talk includes two demos of innovation projects – (i) A parallel-processing ‘data store’ that runs as a scalable in-memory HPC application, hosts different scientific data shapes and orchestrates queries using database-like operators, scalable implementations of theory-inspired algorithms, and pre-trained AI models, and (ii) A Self-Learning Meta-Data Foundation for AI that collects and manages data lineage along with fine-grained metadata generated in the data lifecycle to derive intelligence from the trends observed in the data pipelines. We will demonstrate these projects on scientific workflows in drug discovery, particle tracking, image analysis, etc.
Hewlett Packard Labs
Martin Foltin is a research scientist in AI Research Lab at Hewlett Packard Labs, managing the Data Foundation for AI infrastructure team. Martin has a PhD in Physics and he has been with HPE for 21 years serving in different software and hardware roles, including electronic design automation, VLSI architectures for memory driven computing and AI acceleration, and in AI for Science software infrastructure.
Chief Technology Officer
Hewlett Packard Enterprises
Rangan Sukumar is a Distinguished Technologist at HPE’s Chief Technology Office and the Technical Director for HPE’s AI for Science initiatives. Rangan has a PhD in Artificial Intelligence and was a research scientist and group leader at Oak Ridge National Laboratory, responsible for knowledge discovery and data science workflows on the world’s fastest supercomputers. As a practicing data scientist, his expertise is in handling scientific data shapes that span use cases in drug discovery, autonomous cars, image analysis, and geospatial situational awareness.
HPE AI Team
Microsoft Teams meeting
Join on your computer or mobile app
Click here to join the meeting