What this system does.
In most colleges, student projects are tracked across spreadsheets, email threads and WhatsApp groups. Nobody has the full picture: faculty cannot see status without chasing teams, and students cannot see what is expected of them.
ProjTrack Desk puts the whole project lifecycle in one place, from proposal through milestones to final review. On top of that structure, AI-generated weekly summaries turn raw activity into short, honest progress updates, so faculty can see how every team is doing in minutes instead of hours.
Why the existing setup was not good enough.
Faculty could not answer "how is each team doing?" without chasing people manually. Students did not know what "on track" looked like, and reviewers received PDFs with no context.
How I solved it.
A structured workspace built around the project lifecycle (proposal → milestones → review), with AI-assisted progress summaries that read commit history and standup notes to produce honest weekly updates that faculty actually read.
How it is built.
- 01Django REST backend with role-based permissions for students, mentors and reviewers
- 02React frontend with optimistic UI for status updates
- 03Celery background jobs summarize each team's weekly activity
- 04Notification system routes only the changes that matter
The engineering choices that mattered.
Model the lifecycle, not just tasks
Generic task boards do not know what a proposal, milestone or review is. Building the workflow around the academic project lifecycle means status always means the same thing to students, mentors and reviewers.
Role-based permissions from day one
Students, mentors and reviewers see and edit different things. Enforcing this in the Django REST layer keeps the rules in one place instead of scattering checks across the UI.
AI summaries from real activity
Summaries are generated from commit history and standup notes rather than self-reported status, so the weekly update reflects what actually happened.
Summaries run in the background
Celery jobs generate summaries on a schedule, so model latency never slows down the app and failed runs can retry without anyone noticing.
Notify only on meaningful change
Every update creating a notification is how tools get muted. The notification system filters for changes that need someone's attention, such as a missed milestone or a review request.