A new golden age for data systems is upon us.
Agents consume, manipulate, and rely on state. This state extends beyond databases to files, documents, memory, tools, and the digital environments in which agents act. Supporting agents requires data systems that embrace this broader notion of data and provide new capabilities for managing it. At the same time, agents offer new ways to design new data systems for new application and user needs.
COMS 6113 is a research seminar to study data systems for agents and new semantic workloads, and using agents to support data systems. Topics include but are not limited to:
- Representing and managing agent state
- Supporting agent interactions with data and tools
- Multi-agent coordination and access
- Data safety, identity, and access controls
- Semantic analytics
- Using agents to automate data discovery, analysis, and system optimization
The course will examine usability, reliability, performance, and cost across these settings.
Course Format
Students will study classic data management papers alongside recent research and open questions. The seminar is intended for Ph.D. students and undergraduate or graduate students interested in conducting research. Students should expect to read papers carefully, actively participate and lead in-class discussions, and pursue an ambitious semester-long research project suitable for submission to a top-tier venue in data management, systems, or machine learning. Attendance is mandatory.
Schedule
The meeting time and location are TBD. The schedule below is a tentative outline; dates and readings will be posted before the semester begins.
| Week | Date | Topic | Readings / Due |
|---|---|---|---|
| 1 | TBD | — | |
| 2 | TBD | Readings TBA | |
| 3 | TBD | Readings TBA | |
| 4 | TBD | Readings TBA | |
| 5 | TBD | Project proposals Teams present research questions and plans |
Due: Project proposal |
| 6 | TBD | Readings TBA | |
| 7 | TBD | Readings TBA | |
| 8 | TBD | Mid-semester project check-ins | Due: Progress report |
| — | TBD | Spring break: no class | — |
| 9 | TBD | Readings TBA | |
| 10 | TBD | Readings TBA | |
| 11 | TBD | Readings TBA | |
| 12 | TBD | Readings TBA | |
| 13 | TBD | Readings TBA | |
| 14 | TBD | Final project presentations | Due: Final paper (date TBA) |
Project
The centerpiece of the course is an ambitious, semester-long research project. Projects should be suitable for submission to a top-tier venue in data management, systems, or machine learning.
Project goals:
- Identify a concrete, important research question about data systems for agents or data systems built with agents.
- Situate the question within both classic data management research and recent work.
- Design and build a system, technique, or benchmark, and evaluate it rigorously.
- Present the work to the class and write it up as a full research paper.
Possible directions include:
- Data models, storage, and versioning for agent state and memory
- Interfaces and abstractions for agents interacting with data and tools
- Coordination, isolation, and access control for multi-agent workloads
- Data safety and identity for agents acting on data and tools
- Semantic analytics
- Agents that discover, prepare, and analyze data
- Agents that tune and optimize data systems
Milestones include a project proposal, a mid-semester check-in, a final presentation, and a final paper. Details will be posted before the semester begins.
Syllabus
Course Expectations
- Read the assigned papers carefully before each class.
- Actively participate in discussion, and lead in-class discussions of assigned papers.
- Pursue an ambitious semester-long research project.
- Attend every class. Attendance is mandatory.
Grading
| Component | Weight |
|---|---|
| Paper reading and class participation | TBA |
| Leading discussions | TBA |
| Research project (milestones, presentation, final paper) | TBA |
Attendance
Attendance is mandatory. The topics in this seminar are new and fast-moving, and much of the course's value comes from in-class discussion.
Academic Honesty
Refer to Columbia's academic honesty policy if you are at all unsure. Course-specific guidance on using AI tools will be posted before the semester begins.
FAQ
Who is this course for?
Ph.D. students, and undergraduate or graduate students interested in conducting research. You should be ready to read research papers closely, discuss and debate open questions, and drive a research project.
When and where does the class meet?
The time and location are TBD and will be posted on this page.
Is attendance required?
Yes. Attendance is mandatory.
What kind of project will I do?
An ambitious semester-long research project on data systems for or with agents, aimed at submission to a top-tier venue in data management, systems, or machine learning. See Project.