CS 5416

Cloud Computing and ML Hosting

Course Description

Course information provided by the 2025-2026 Catalog.

CS 5416 is a course focused on the systems aspects of performance for cloud-hosted AI and ML applications such as LLMs, as well as complex systems in which AI or ML is just one element. The T/R lectures are shared with an undergraduate class, CS 4414 but the (required) recitation sections cover content that the CS 4414 students will not see and the last of the multi-stage projects will focus on cloud AI and ML scenarios, while the last of the CS 4414 projects focuses on a single-computer scenario. Thus even though you will overlap with Cornell seniors (and some juniors) in the lecture hall, you will also gain additional perspective and knowledge in recitation and will be doing a more ambitious style of project that comes very close to what companies are undertaking to integrate AI into their cloud solutions. The CS 4414 students will have a separate discussion board than you, and although you do take the same prelim exams, your grading is more focused on projects (25% exams, 75% projects, in contrast to the 50-50 balance for CS 4414). CS 5416 starts by exposing students to programming applications at the systems level and to the operating systems abstractions that these applications depend on. It then builds on this foundation to look at systems issues that shape the performance and reliability of modern ML and AI applications, such as “chat bots” and question-answering AIs. We do not expect students to learn how these ML and AI tools work, in a mathematical sense. Instead our focus is on how they execute, where components run and how they talk one another, how they interact with big-data storage, and how they leverage accelerators such as GPU. CS 4414 students who attend these cloud-computing and ML lectures won’t be using the material in their projects, whereas CS 5416 students will gain hands-on experience working with and optimizing ML and AI applications that have this form. The required recitation is where you will learn about the additional tools and techniques that are needed to perform those tasks

Last 3 terms offered (None)

Learning Outcomes REF-FA25

• Background: Proficiency programming in C++ 20, using Visual Studio Code IDE on Linux (Ubuntu) demonstrated through successful completion of hands-on assignments. Proficiency with Linux commands and bash programming demonstrated through successful completion of hands-on assignments. Ability to write multithreaded code that leverages the full performance of modern NUMA servers demonstrated through successful completion of hands-on assignments that focus on speeding up code by using multicore parallelism. 6
• Understanding performance in cloud-hosted microservice systems with AI and ML components. Interpretation of parallelism in many forms, and ability to create parallel solutions to practical computing problems, to implement them correctly in C++, and to debug and optimize solutions. Demonstrated through a mix of exam performance and ability to use these ideas when creating hands-on assignments.
• The concrete skill to actually write code in C++ that performs well in these cloud microservice settings. This departs from what the CS 4414 students will be learning (their focus is on programs running in a single Linux server, and often performing systems tasks rather than AI-assisted tasks).
• The ability to diagnose performance issues and bugs in microservices that have AI or ML components, RAG databases, or other kinds of data repositories. Again, this departs sharply from what the students in CS 4414 will be learning. The additional knowledge and skills will be taught in recitation, which is why we require you to attend those as well as the main (shared) lectures.

View Enrollment Information