UCLIC Research Seminar Series
In recent years, data science has emerged as one of the most exciting disciplines for facilitating data-driven decision-making in various domains, from banking and finance to agriculture and healthcare. Data science workflows are human-centered processes where stakeholders interact with different tools and interfaces to accomplish their tasks — from spreadsheets and visualization dashboards to scripts and computational notebooks. Therefore, data systems that support these workflows need to optimize for diverse objectives (e.g., usability and interactivity), stakeholders (e.g., data scientists and product managers), and domains (e.g., text, image, and audio.) In this talk, I will present takeaways from a study of data practitioners that informs design guidelines for building human-centered data systems. I will discuss the implications of instrumenting these guidelines in tools for diverse tasks such as exploratory text analysis and graph data science. Finally, I will discuss how lessons learned from these experiences highlight multifaceted challenges spanning human-centered design and data management for data science. The primary message of the talk is that these challenges are interconnected and cannot be addressed in isolation. Rather a holistic effort spanning different disciplines (e.g., DB, AI, and HCI) is required when exploring various aspects — such as interface design, data modeling, and infrastructure development — that support data science workflows.
*This seminar is an online event via Zoom (https://ucl.zoom.us/j/97671517328)
Sajjadur Rahman is a Senior Research Scientist at Megagon Labs, Mountain View, California, where he leads the human-centered data science research group. Before joining Megagon Labs, Sajjadur obtained his Ph.D. in Computer Science at the University of Illinois at Urbana-Champaign. Prior to that, he served as a Lecturer at the Department of Computer Science and Engineering at Bangladesh University of Engineering and Technology. Sajjadur's research focuses on building human-centered data systems that are interactive, explainable, and scalable by synthesizing techniques from databases, visualization, and HCI. In particular, his research explores various approaches to enhance the productivity of data practitioners --- from building interfaces for effective human-agent collaboration to developing algorithms that facilitate large-scale data analytics. His research has been published in premier conferences in Databases (SIGMOD and VLDB), HCI (CHI and CSCW), and NLP (EMNLP and NAACL.) His work has been recognized with the best demo award (at ICDE) and featured in popular tech blogs. His research identified performance bottlenecks of popular spreadsheet systems such as Microsoft Excel and Google Sheets, enabled scalable data exploration in open-source tools (e.g., the Berkeley/UofI DataSpread Project), and enjoyed industry adoption (e.g., a Human-AI collaborative task management tool at B12.) Sajjadur has organized workshops and served as panelist and PC member at conferences such as SIGMOD, VLDB, ACL, and IEEE ISSRE.