Writing
Essays on the parts of teaching computer science I think about most: how to assess understanding when AI tools can do the homework, and how to teach data engineering with the systems and constraints that real work runs on. Most of these started as talks or course materials.
My teaching philosophy in one place: connect rigorous theory to real industry practice, integrate AI into learning and assessment without surrendering the thinking to it, and keep every student in their zone of productive struggle. The piece that frames the other three.
Once a chatbot can write a passable answer, a written exam stops measuring understanding. Oral exams still do. How I run a design-interview-style oral exam, grade it quickly, scale it with TAs to large classes, and why it costs about the same as a written exam.
A grade should require both mastery and effort. A weighted sum lets one paper over the other, which is the loophole AI exploits. So multiply the components instead of adding them. The design, the reasoning, and results from a 2024-25 implementation.
A CSE 5114 assignment: build a two-tiered streaming pipeline over Wikipedia's live edit stream with Kafka, Flink, and a rate-limited language model, where a cheap filter escalates only ambiguous edits to an expensive model. A real infrastructure archetype, on live data.