Sajjadur Rahman
444 Castro St, Mountain View, CA 94041
I am a Senior Research Scientist at Megagon Labs and the founding Research Manager of the Data-AI Symbiosis (DAIS) group. I received my PhD from CS@Illinois where I worked with Aditya Parameswaran. Before that, I was a lecturer at Department of CSE, BUET.
My research synthesizes techniques from data management, AI, and HCI to build scalable, interactive, and responsible systems for data discovery and exploratory analysis. Currenly, my focus is on building a data platform for LLM-powered multi-agent systems while prioritizing aspects such as fact checking and verification, text-2-DSL querying, integrating information within data lakes, and data lakes usability. My research has been published in premier conferences in Databases (SIGMOD and VLDB), HCI (CHI and CSCW), and NLP (EMNLP and NAACL.) I collaborated on research projects that were recognized with the best demo award (at ICDE) and featured in popular tech blogs. My research enabled scalable data exploration in open-source tools (e.g., DataSpread), and enjoyed adoption in the industry (e.g., a planning tool for creative design at the startup B12.) I was the Program Chair of the MATCHING workshop at ACL 2023 and served on the program committee of SIGMOD, VLDB, EMNLP, and IEEE ISSRE.
News
- Paper on "Characterizing LLMs as Rationalizers" accepted at ACL'24 Findings.
May 15, 2024
- Paper on "Benchmarking Data Discovery in the Enterprise" accepted at GUIDE-AI@SIGMOD'24.
May 12, 2024
- Demo on "Human-LLM Collaborative Annotation System" accepted at EACL'24.
January 22, 2024
- Paper on "Human-LLM Collaborative Annotation Through Effective Verification" accepted at CHI'24.
January 18, 2024
- Received recognition for outstanding review at CHI 2024.
November 15, 2023
- Founded the DAIS group to lead our efforts in building data platforms for multi-agent systems.
November 01, 2023
- Organized the MATCHING Workshop@ACL'23.
July 13, 2023
- Delivered a talk at UCLIC research seminar.
February 15, 2023
We observed that data science workers follow an iterative task model consisting of information foraging and sensemaking loops across all the phases of an information extraction workflow. We found several limitations in both loops stemming from a lack of adherence to existing cognitive engineering principles.
Leam is an interactive tool for text data analysis that, backed by a visual text algebra, combines the strengths of spreadsheets, computational notebooks, and visualizations libraries. The visual text algebra supports a number of text analysis and visualization operations.
IncVisage is a progressive visualization tool that reveals “salient” features of a visualization quickly while minimizing error, enabling rapid and error-free decision making. The approach is orders of magnitude faster than the traditional visualization systems.