Dr. Amir Shaikhha
Title: Democratizing Data Science by Leveraging Structure
Abstract: Modern data science pipelines have become increasingly complex, employing diverse workloads such as tensor algebra, graph processing algorithms, and relational query processing. This diversity often leads to the use of loosely coupled data processing frameworks that require moving data across different stages of the analytics pipeline. However, this constant movement of data introduces significant inefficiencies in resource utilization and increases energy consumption—critical concerns in today’s computing environments.
In this talk, I present a novel compilation-based approach designed to bring computationcloser to the data, thereby mitigating these inefficiencies. This method focuses ondesigning domain-specific languages (DSLs) that leverage the data’s inherent structurethrough algebraic optimizations. By tailoring these DSLs to specific data types andcomputational patterns, we can perform computations more efficiently and reduce theneed for excessive data movement across the pipeline. I will demonstrate how thisapproach significantly outperforms state-of-the-art frameworks across a wide range ofapplications, including database query processing and tensor operations, illustratinghow compilation techniques and domain-specific languages can be harnessed tooptimize data-intensive workloads and pave the way for more efficient and sustainabledata science practices.
Bio: Amir Shaikhha is an Associate Professor (Reader) in the School of Informatics at theUniversity of Edinburgh. His research focuses on the design and implementation ofdata-analytics systems by using techniques from the databases, programminglanguages, compilers, and machine learning communities. He was a DepartmentalLecturer at the University of Oxford (2019-2020) before starting as an AssistantProfessor (Lecturer) at the University of Edinburgh (2020-2024). He earned his Ph.D.from EPFL in 2018, for which he was awarded a Google Ph.D. Fellowship in structureddata analysis, as well as a Ph.D. thesis distinction award. He has won the Best PaperAward at GPCE 2017, the Most Reproducible Paper Award at SIGMOD 2017, and theMost Influential Paper Award at GPCE 2024. He (co-)chaired the program committeesof DBPL 2021, Scala 2022, DRAGSTERS 2023, GPCE 2023, and Sparse 2024.
Dr. Mohammad Bakhshalipour
Title: Rethinking Architectures for Robotics
Abstract: As robotics technology becomes a major driver of future societal advancements, withforecasts predicting the deployment of 20 million units and a market valuation of $70billion by the end of this decade, it is essential to rethink the computer architectures thatsupport these systems. Current processors are not equipped to meet the real-time,high-performance demands of robotics, exposing a significant disconnect between thefields of robotics and computer architecture. This talk highlights the need forarchitectural innovations in two key areas: (i) the development of comprehensive, open-source benchmark suites for robotic applications, and (ii) the design of efficient,robotics-specific computer architectures. These benchmark suites provide crucialinsights into the limitations of current hardware, emphasizing the need for new tools thatcan guide future research. Additionally, the talk introduces novel architectural solutionstailored for robotics, including application-specific hardware accelerators and domain-specific processors, which address key performance bottlenecks in robotic operations.
Bio: Mohammad Bakhshalipour received his PhD in Electrical and Computer Engineering from Carnegie Mellon University, where his thesis, “Bridging Robotics and Computer Architecture,” was awarded the A. G. Milness Award. His research focuses on the intersection of computer architecture and robotics. He holds an MSc and BSc in Electrical Engineering from Sharif University of Technology, obtained in 2017 and 2015, respectively.
