CMSE 382: Optimization in Data Science

For the fall semester of 2025, I served as the Graduate Teaching Assistant for CMSE 382: Optimization in Data Science. This is a high-level course primarily consisting of juniors and seniors in the BS in Data Science program in CMSE. I worked under Dr. Firas Khasawneh with a 30-to-1 student-instructor ratio.

I was primarily responsible for grading assignments, facilitating class time with Dr. Khasawneh, and holding weekly office hours, although at times I did create content upon his absence.

Grading

The class had 59 students enrolled in the course, and I graded four kinds of assignments:

• In-class, daily worksheets

  • • These were graded for effort and submission on a 0-1-2 scale. In the partially-flipped classroom environment, the students worked on worksheets during the last half of class everyday. My average time to return grades was less than one business day.

• Biweekly homework assignments

  • • These assignments consisted of a few problems and were graded for accuracy. I was given the amount of points per problem and was then responsible for creating rubrics for the assignments to ensure partial credit was granted and equitable. Comments were frequent when relevant and were specific to provide structured and actionable feedback. The average time to return grades was between 2 and 3 business days.

• Biweekly short quizzes

  • • This was another type of summative evaluation graded for accuracy that occurred on offsetting weeks from the homework assignments. These consisted of one to two free response questions and then a series of multiple-choice questions. Similarly, I was responsible for creating rubrics to align with the learning objectives we were testing while maintaining transparent grading. My average grade release occurred less than two business days following the quiz.

    • Additionally, I would take the quiz myself before distribution to ensure the questions were well-defined and achievable, giving myself half the time of the students.

• 3 Midterm examinations distributed evenly across the semester

  • • The midterms were the only assessments where grading duties were split between myself and Dr. Khasawneh. Each test consisted of 7 open-ended questions, where I would grade four of them. Again, I created well-structured rubrics that aligned with the learning objectives and distributed percentage weights of concepts and partial credit to match that of the quizzes and homework assignments. My average time to grade was ~3 days, but frequently grades were returned later due to makeups and other factors.

    • Similarly, I took each of the tests myself to ensure the questions were clear and manageable.

All grading was done through Crowdmark, an online tool to assist in streamlined grading that syncs well with D2L/Brightspace that MSU uses. Comments were actionable and clear, taking advantage of Crowdmark's ability to use Latex, and frequently color-coded to sections of their work. The responsibility of the creation of assignments on Crowdmark alternated between the two of us

Class Facilitation

Students were divided into 10 groups to work together on in-class assignments. These groups were randomly assigned and changed once about midway through the semester.

The first half of class consisted of a few minutes of announcements and reminders and the rest lecture on that day's material. We would both give announcements and reminders about upcoming assignments and such, and then I would take attendance based on the groups. At the beginning of the semester, attendance would have to wait until I could walk around without disruption, but it only took me about two weeks to be able to learn names and faces to do the attendance silently.

Once Dr. Khasawneh had finished lecturing, he would signal me to release the in-class assignment and then we would walk around answering questions. When students asked me questions, I would prompt them to rephrase it into an actionable question that could be answered and focused on the concepts and processes rather than the solution. Additionally, my first reply would often be asking if they have asked their group members, reminding them the groups exist for a reason.

I have found a reasonable balance between maintaining professionalism while promoting approachability, frequently talking to them about their interests and stressing the broad applicability of the course material to their future careers.

Office Hours

I held two 90-minute sessions weekly for office hours with additional remote hours available by appointment. My office hours were relatively well attended with roughly 20% of students attending at times. I kept track of questions and would then match the questions to learning objectives. This data informed the review days before the midterm exams as well as served as notes for improvement for future iterations of the course.

Some students would come with specific questions in mind, and some would attend and simply work on their homework and ask questions as they arose. I would offer study tips, would go over guided problems on the whiteboard with them, and would rephrase the key theorems and concepts in plainer or alternative terms.

A few students couldn't attend either of my sessions and would email me questions or schedule a virtual one-on-one. My additional availability was limited, but when able, I would meet with them over zoom to go over their questions and guide them on their understanding of the concepts.

Content Creation

Dr. Khasawneh was away on travel for over a week of class twice throughout the semester. For the first instance, he was able to get other faculty members to cover him except for one class which I covered. Content was created for the material by the professor in advance, but I also created material to stress the different roles between myself and Dr. Khasawneh and when to email whom.

One piece of content I created for this class period was to gather information on how the students chose to prepare for class. This helped inform Dr. Khasawneh which content and media to prioritize in creating and inform the both of us in interacting with students in problem-solving. Frequently, as I would go around in class and see the students with various AIs pulled up, I would tell them that the book would be much more helpful and guaranteed to be accurate; this was more effective since I knew most of them didn't read the book.

Additionally, Dr. Khasawneh was away for four class periods in a row later in the semester leading up to a midterm. He had prepared in-class assignments and the slides for the new content we would cover, but the majority of the days were review days. Based on interactions and requests from the students, I created content that was designed to target those needs. Some of them were worries about knowing the material well enough but not being able to do it in a short timeframe, so I created content designed on building intuition and speed. Another common worry was the quantity of theorems and other concepts we were covering on the exam, so I created an abridgement of the important theorems and definitions to help focus their time in a few different forms, namely a compact document of the material and then the material phrased as multiple-choice questions. Of course, the students knew there were only free-response questions on the exam, but this allowed them to test their knowledge of the material concretely and quickly.
The creation of this review material allowed students to study their weakest link led to a tangible improvement compared to midterm 1, even though the material was harder and required knowledge from the first midterm. Specifically, the mean increased 7 points, the median increased 1.5 points, and the standard deviation decreased almost 7 points. As shown, the creation of this review material lifted the grades of the overall class, but more substantially improved the students with the worst performance.

Comments on how did the course support my learning:

"Cole also created additional material in response to student feedback to make it easier for us to learn, and overall broke down the material so that it was easy to learn and apply. I'm not a very math-oriented person, but this was probably one of my favorite CMSE courses so far. Cole was also one of the best TAs I've ever had and he really stepped up when Dr. K had to miss some lectures."