{"677868":{"#nid":"677868","#data":{"type":"event","title":"Ph.D. Proposal Oral Exam - Po-Kai Hsu","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u0026nbsp; \u003C\/strong\u003E\u003Cem\u003EMemory Centric Hardware Accelerator Design using 3D Memories For Ultra-large AI Models\u003C\/em\u003E\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ECommittee:\u0026nbsp;\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EDr.\u0026nbsp;Yu, Advisor\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EDr. Khan, Chair\u003C\/p\u003E\u003Cp\u003EDr. Datta\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EThe objective of the proposed research is to address the inefficiency of conventional Von Neumann architectures in processing the ultra-large artificial intelligence (AI) models with memory-centric hardware accelerator. The exponential growth of data in genomics and AI has highlighted the limitations of conventional Von Neumann architectures, particularly their inefficiency in handling data-intensive applications. These systems suffer from high energy consumption and latency due to excessive data movement between memory and processing units. To address these challenges, this research proposes memory-centric hardware accelerator designs using 3D NAND Flash and 3D DRAM technologies. The goal is to develop novel architectures that leverage in-memory and near-memory computing to minimize data movement and improve energy efficiency and scalability. The proposed designs include a 3D NAND Flash-based hyperdimensional computing (HDC) engine for genome sequencing and mass spectrometry, and a 3D DRAM-centric accelerator for large language model (LLM) fine-tuning. By performing computations directly within or near memory arrays, these architectures aim to overcome the computational bottlenecks of traditional systems. Simulation results will benchmark the proposed designs against existing GPU-based accelerators, focusing on key metrics such as energy consumption, latency, memory density, and computational throughput. This research contributes to the development of efficient and scalable memory architectures for data-intensive applications in genomics and AI.\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Memory Centric Hardware Accelerator Design using 3D Memories For Ultra-large AI Models"}],"uid":"28475","created_gmt":"2024-10-23 16:54:55","changed_gmt":"2024-10-23 16:56:06","author":"Daniela Staiculescu","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2024-10-28T10:00:00-04:00","event_time_end":"2024-10-28T12:00:00-04:00","event_time_end_last":"2024-10-28T12:00:00-04:00","gmt_time_start":"2024-10-28 14:00:00","gmt_time_end":"2024-10-28 16:00:00","gmt_time_end_last":"2024-10-28 16:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Online","extras":[],"related_links":[{"url":"https:\/\/teams.microsoft.com\/l\/meetup-join\/19%3ameeting_MWEzOWFlNDQtNTM0Yi00MjhiLThjYTMtYTgxN2U2NzQ0NWQw%40thread.v2\/0?context=%7b%22Tid%22%3a%22482198bb-ae7b-4b25-8b7a-6d7f32faa083%22%2c%22Oid%22%3a%22727e4461-23c7-4abf-96e4-33aa07526752%22%7d","title":"Microsoft Teams Meeting link"}],"groups":[{"id":"434371","name":"ECE Ph.D. Proposal Oral Exams"}],"categories":[],"keywords":[{"id":"102851","name":"Phd proposal"},{"id":"1808","name":"graduate students"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}