SPRING 2027 Senior Project Descriptions
Professor: Michael Grossberg
Multi-Agent AI Systems: Designing, Building, and Evaluating Teams of LLM Agents
Large language model agents can now call tools, write and run code, search the web, and coordinate with one another. Whether a team of agents works better than a single agent or an ordinary program is still an open question. In this two-semester project course, teams of 3 to 5 students design and build an original multi-agent system. Each team picks a real problem, such as NYC civic data, software maintenance, accessibility, education, or data analysis. The team must show why the problem needs more than one agent, and must measure whether it does. In the first semester (CSc 59866), I will teach what students need to get started: how LLMs and tool-calling agents work, how to build an agent loop from scratch, how multi-agent systems are organized and how they fail, how to evaluate agents, and how to keep them secure. We will also practice professional software engineering with Git workflows, code review, automated testing, reproducible environments, and continuous integration. Each team will survey existing systems and research, state clearly how its project differs from them, and build a small prototype of its riskiest idea. The semester ends with a written proposal and an oral defense.
In the second semester (CSc 59867), teams carry out their approved proposals through regular status meetings and milestones. These cover baselines, a working end-to-end system, testing, evaluation, a midpoint demonstration, and a final demonstration, report, and presentation. Projects should be original and open source, and each must include a quantitative evaluation. Projects that only wrap an existing API or reproduce a published system will not be approved.
Section-Specific Notes:
• I prefer students who have already taken the CCNY Machine Learning and/or Artificial Intelligence courses, and who have hands-on experience using LLMs. These are not required. Students without them should expect a lot of catching up on their own.
• The course uses Python, NumPy, pandas, PyTorch, and Matplotlib throughout. I will not teach them. Students should be comfortable with them before the semester starts.
• The course cannot provide computer resources. Students should plan to use their own computers and low-cost or free tools, and should consider this when proposing a project. Projects must be feasible within those limits. Projects that need large-scale computing are poor fits.
Professor: George Wolberg
In this project-based course, students are grouped into teams of up to three students per group to work on building apps for iOS, Android, web, and Windows/Mac platforms. The capstone course will last two semesters. In the first semester, we will study how to build, test, and deploy mobile, web, and desktop apps from a single codebase. After fundamental principles are introduced, each team chooses one topic of interest from the fields of image processing, computer vision, computer graphics, and gaming. Groups will perform research and development to specify deliverables, milestones, and implementation considerations. Each group must research similar apps in their chosen topic and resolve how their own app will distinguish itself from the available offerings, present these findings to the class, demonstrate a deep understanding of the principles and algorithms that the app will implement, and outline a working plan to implement the software. In the second semester, each team will focus on implementation and presentation of ongoing results with detailed design, implementation, integration, testing, and experiment evaluation. Projects will be selected based on the interests of the students and professor.
The goals of the course are to gain deeper insights into the workings of real-world software engineering, to receive valuable hands-on experience in basic research, and to better understand how to build, test, and deploy mobile, web, and desktop apps from a single codebase.
Instructor: Nikolai Avteniev
AI Engineering Senior Project Course Description:
This two-semester Senior Project sequence focuses on the engineering of software systems that incorporate large language models (LLMs) and generative AI as core components. Students will learn how LLMs work and how to apply techniques such as prompting, retrieval-augmented generation, tool use, agents, evaluation, and deployment. The course emphasizes challenges unique to LLM-based systems, including nondeterminism, model selection, system architecture, observability, cost, security, and responsible use. In Senior Project I, students will build the technical foundation needed to design these systems and, working in teams, identify a problem that depends on LLM or generative-AI capabilities, evaluate alternative approaches, and produce a detailed project proposal and system design. In Senior Project II, teams will implement, test, evaluate, and iteratively refine their proposed systems based on evidence gathered during development.
Last Updated: 10/05/2026 19:54