A certificate is only as credible as the assessment behind it. This post describes how the AI Engineering Bootcamp measures competency, and why we chose continuous, evidence-based assessment over a single final exam.
The problem with attendance-based credentials
Many programs award a certificate for finishing. That tells an employer that a student was present, not that the student can build anything. Our academic standards require the opposite: every one of the sixteen required modules is graded, students must pass each term to advance to the next, and the results are recorded on an official transcript.
Five instruments, one transcript
1. Weekly graded assignments. Every week from week 1, students submit working code. Submissions pass through automated correctness checks first, then receive human instructor feedback on design, readability, and engineering judgment. Both layers are graded.
2. Practical coding assessments. At the midpoint and end of each term, students complete timed, hands-on assessments: building or debugging a real system, not answering multiple-choice questions. A Term 2 assessment requires implementing a working RAG pipeline with chunking, embedding, retrieval, and structured outputs within the time limit.
3. Automated AI technical evaluations. For LLM systems, correctness is not binary, so we grade the way production teams do: automated evaluation harnesses score each student's system for answer quality, retrieval precision, latency, and token cost against a published rubric. Students see their evaluation traces - the same LangSmith traces they will use professionally.
4. Structured peer review. Each week, students review classmates' pull requests. The reviews themselves are graded for technical accuracy and communication quality, because code review is an assessed learning outcome, not a courtesy.
5. Projects and the capstone. The midterm project (an individually built, end-to-end LLM application) and the team capstone are graded by faculty panels against published rubrics, including an architecture review and a final technical defense.
What "competency-based" means in practice
A student who cannot demonstrate a required competency does not advance, regardless of attendance. Remediation exists (office hours, mentoring, and reassessment windows are built into every term), but the standard does not move. That means we can publish our learning outcomes as claims about every graduate rather than aspirations about some.
The transcript
Every assessment above appears on the graduate's official transcript: module grades, assessment results, project evaluations, and the capstone defense outcome. Employers may request transcript verification for any graduate through the admissions office.
The complete assessment policy is published in Section 06 of the Program Details.