{"693633":{"#nid":"693633","#data":{"type":"event","title":"Ph.D. Proposal Oral Exam - Ritik Raj","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle:\u0026nbsp; \u003C\/strong\u003E\u003Cem\u003EWorkload-Aware Hardware-Software Co-Design for Efficient Agentic AI on Heterogeneous CPU-Accelerator Systems\u003C\/em\u003E\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ECommittee:\u0026nbsp;\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EDr.\u0026nbsp;Krishna, Advisor\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003Cp\u003EDr. Mahajan, Chair\u003C\/p\u003E\u003Cp\u003EDr. Hao\u003C\/p\u003E\u003Cp\u003EDr. Huang\u003C\/p\u003E\u003Cp\u003EDr. Wang\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003EThe objective of the proposed research is to improve the end-to-end efficiency of agentic AI systems through workload-aware hardware\u2013software co-design. These systems combine large language model (LLM) inference on accelerators with tool execution on CPUs. CPU processing can dominate tool-intensive workloads, while LLM-intensive workloads introduce challenges in model routing, load balancing, and memory management. For LLM-intensive agents, ORACLE uses task-specific verification and asynchronous feedback to improve model selection. Its scheduler accounts for KV cache capacity and balances the benefit of shorter waiting times against potential losses in accuracy. For tool-intensive agents, OneDSE uses workload information to predict CPU configurations and simulation feedback to guide further search, reducing the number of evaluations needed for design exploration. The thesis also presents a cycle-accurate simulator for exploring accelerator designs, along with dataflow optimizations that reduce memory traffic in symbolic reasoning. The proposed research further investigates mixed-precision acceleration based on each model\u2019s role in an agentic program and joint routing and scheduling for model pools that change with the workload. Together, these efforts aim to improve performance and resource utilization across heterogeneous CPU\u2013accelerator platforms by coordinating hardware capabilities with the computational and memory demands of agentic execution.\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Workload-Aware Hardware-Software Co-Design for Efficient Agentic AI on Heterogeneous CPU-Accelerator Systems"}],"uid":"28475","created_gmt":"2026-10-29 19:23:50","changed_gmt":"2026-10-29 19:25:57","author":"Daniela Staiculescu","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2026-11-18T10:00:00-05:00","event_time_end":"2026-11-18T12:00:00-05:00","event_time_end_last":"2026-11-18T12:00:00-05:00","gmt_time_start":"2026-11-18 15:00:00","gmt_time_end":"2026-11-18 17:00:00","gmt_time_end_last":"2026-11-18 17:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Online","extras":[],"related_links":[{"url":"https:\/\/teams.microsoft.com\/meet\/2608173521748671?p=CXSU5Ba8adeIe4Nuo3","title":"Microsoft Teams 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":""}}}