Capstone projects
Study-first. Engineering is a floor. Every project produces a defensible empirical claim.
Orientation
The course’s final project option (0–30 points + 3 bonus) is study-and-analysis first. Teams of 1–3 students build the smallest artifact that lets a real empirical study run end-to-end, then invest the freed time in sharper questions, cleaner measurement, and honest statistics. Polish beyond the minimum build is flagged as engineering performed in place of science.
Working documents
Candidate capstone topics: conversational clustering, graph-based data generation, AI happiness, conversational graph extraction.
The ten components every project must include — organised in three parts: study design, just-enough engineering, analysis & reporting.
What every project ships, how sprints are run, and the two-pass grading rubric that determines the final project mark.
Three dimensions every project must cover
- Learning — real related-work survey and methodological grounding; never the deep deliverable on its own.
- Building — a minimum viable working artifact. Instrumental to the study, not the product.
- Experimentation & analysis — the depth of the project. Hypothesis-driven, versioned study plan, proper statistics, claim with quantified uncertainty.
See Universal Scaffolding for the authoritative component list and Deliverables & Rubric for how the final grade is assigned.