AI-assisted academic operations system

Personal automation project · Claude as an AI agent with browser and desktop automation

Claude (AI agent)Canvas REST APIJavaScriptBashAppleScriptGmail searchiCalmacOS automation

The problem

Four online courses across two learning platforms (Canvas plus vendor lab portals), with broken assignment links, scores that didn't sync, misconfigured grade totals, and instructor communication spread across three inboxes. There was no single view of what was outstanding or what could still be recovered.

What I built

An AI-assisted workflow that audits every course, prioritizes recovery work, tracks instructor follow-ups, and sends a daily status report by text.

1. Full course audit

Connected to the LMS through an authenticated browser session and queried its REST API for every course: assignments, submissions, assignment groups, and grading weights.
Found missing work the platform's own "missing" view didn't show, by comparing due dates, submission status, and zero scores.
Rebuilt actual grades from syllabus weights, because the platform's totals were wrong: it counted future assignments as zeros and had broken category weights.

2. Policy-based triage

Pulled late-work policies from each syllabus (PDF and DOCX) and sorted every outstanding item as recoverable, needs approval, or closed.
Ranked recovery work by points at stake and deadline risk, then built a time-blocked plan to clear it.
Calculated the minimum score needed on remaining work for each course.

3. Email and communication tracking

Searched multiple inboxes with targeted Gmail queries (sender, subject, date range) to find instructor requests and support threads.
Drafted concise follow-ups to instructors about grade-sync problems and broken links, with evidence attached.
Surfaced time-sensitive items buried in school email, such as a paid internship notice.

4. Platform troubleshooting

Diagnosed lab scores that weren't syncing to the LMS (LTI launch problems) and documented them for instructors.
Tracked down missing courseware access, and found proof of completed work on disk, such as a certificate that had never been uploaded.

5. Automated daily brief

A scheduled automation compiles grades, deadlines in the next 3 days, outstanding work, new grades, and important emails into one text message.
A no-login fallback through a private iCal feed keeps due dates coming when sessions expire.
Delivered by iMessage through AppleScript. Diagnosed and fixed duplicate delivery caused by a dual-line phone setup.
Security practices: read-only access, secrets stored locally with restricted permissions, nothing sensitive in logs.

Results

Fully caught up across all courses in about 2 days.
Recovered 150+ points through accepted late work, resubmissions, and instructor grade corrections.
Brought one course up to a B average.
Replaced manual checking across 3 platforms and 2 inboxes with one automated daily text.

← back to portfolio