THURSDAY (MAY 23, 2024) – ISEB 1010 @ UCI
Building a Feature Store for Personalized Search/Recommender System
Speaker: Yiming Ma
Abstract: After graduating from ICS in 2007, I had opportunities working in a research lab, startups, mid and large tech companies. The knowledge and experiences I acquired during my time in the database group, under the guidance of Professor Mehrota and Professor Li, have proven invaluable to both my work and research endeavors. In this talk, I will have the privilege of sharing insights gained from my past work on Building Feature Stores for Personalized Search/Recommender Systems. Throughout the session, I will delve into the evolution of feature representations over the years and the corresponding advancements in underlying systems. This exploration will encompass the ever-growing demands of modern AI modeling and inference, with a particular emphasis on large-scale implementations. Considering the vast scope of this subject matter, I acknowledge the limitations of my own experiences. Nevertheless, I am excited to reconnect with the brilliant minds at ICS, as this talk serves as an opportunity for us to exchange knowledge and learn from one another.
Bio: Dr. Yiming Ma is an accomplished professional with a strong academic background and extensive industry experience. He obtained his doctoral degree from the esteemed database group at UC, Irvine in 2007, working under the guidance of Prof. Mehrotra and Prof. Li. During his academic tenure, His primary research focus revolved around aiding first responders in developing spatial situational awareness across diverse scenarios. Building upon his expertise, Dr. Ma transitioned into his current role as an AI platform manager at Axon Enterprise, where he continues to contribute to the advancement of building situational awareness for the law enforcement communities. Prior to his role at Axon, he held different R&D positions, including that of an Ex-Googler and Sr. AI Tech Lead & Manager at LinkedIn. Throughout his career, Dr. Ma has played roles in leading and contributing to multiple mission-critical end-to-end cross-functional AI platform projects. His areas of expertise encompass a wide range of subjects, including Feature Store construction, offline model training, and online model deployment and serving. This extensive knowledge and experience have positioned him as a solid data management professional in the field.
(Healthcare) Data Mapper
Speaker: Shengyue Ji
Abstract: Healthcare data is highly complex and fragmented. The industry has spent years moving towards FHIR (Fast Healthcare Interoperability Resources). In this work we designed, implemented, and launched a product called Data Mapper that helps data scientists with clinical expertise to quickly drive the data mapping process, using basic knowledge of SQL. This low code product has been proven to work for mapping many EHR (Electronic Health Record) data, decreasing the task timeline by at least 4x. It could also be easily extended to serve other industries’ needs.
Bio: Dr. Shengyue Ji is currently a Software Engineer at Google working on healthcare products and systems. Previously he was at Uber working on real time analytics. Before that he was at Google working on Google Analytics and Google Tag Manager. He earned his PhD under the guidance of Professor Li in Information and Computer Science from University of California, Irvine.
Leveraging Ph.D. Insights in Industry: A Journey Through Five Key Experiences
Speaker: Ronen Vaisenberg
Abstract: This talk explores the practical applications of Ronen Vaisenberg’s Ph.D. training in his professional journey at Google. Highlighting five pivotal experiences, Ronen discusses how SQL provided critical product insights, database theory shaped his problem-solving intuition, and statistics drove real-world solutions. He shares the impact of effective presentations in aligning stakeholders and emphasizes the broad utility of Ph.D.-acquired skills as a safety net. Attendees will gain insights into bridging academic research with industry practices, emphasizing the value of a Ph.D. in navigating complex technical challenges and fostering innovation.
What Do You Want to Be?: A Journey of Reinvention and Joy
Speaker: Yun Huang, UIUC
Abstract: “What do you want to be?” was the question my Ph.D. advisor asked me on the first day of my doctoral program. In this talk, I will share my journey of reinvention and personal growth as I navigated various roles in academia and industry. I will recount stories of staying true to my non-negotiables, how collaborating with interdisciplinary teams has shaped my research, and how I transitioned to new scholarly communities. These experiences taught me the importance of simplicity in my priorities, openness to new adventures, and finding joy in my endeavors. I will reflect on the challenges faced, the lessons learned, and the moments of clarity that have defined my path and aspirations.
Bio: Dr. Yun Huang is an Associate Professor at the School of Information Sciences at the University of Illinois at Urbana-Champaign. She is dedicated to innovating AI-based solutions that cultivate a synergistic relationship between humans and machines, enhancing educational opportunities to all, and increasing access to community services. She is deeply committed to translating research into real-world benefit and making a positive societal impact. Her work has garnered support from respected government agencies such as the National Science Foundation, the Institute of Museum and Library Services, and the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR) and leading industry names, including OpenAI, Google, IBM Research, and HeyGen. She earned her Ph.D. in information and computer science from the University of California, Irvine.
Title: Spark/Photon Unsolved Challenges
Speaker: Alex Behm
Abstract: In this talk, I will introduce you to the Databricks data processing stack. You will learn about the concept of a Lakehouse, Spark, as well as Photon, the vectorized execution engine. I will take you on a tour of challenging customer scenarios that stretch the boundary of what Databricks can handle. As we will see, even seemingly simple queries become interesting when faced with extremes. I hope to pick your brain on practical solutions!
Bio: Alex has been building databases for over a decade in academia and industry and maintains a passion for speed and quality. He is the tech lead for Photon, a new vectorized engine written from scratch in C++ that powers Databricks. Before joining Databricks, Alex worked on Apache Impala as the second engineer on the project. Alex holds a PhD in databases from UC Irvine.