AI AUTOMATED ATTENDANCE PIPLINE

AI-powered attendance monitoring that detects risk early and automates student intervention. Takes automated attendnce and monitor student no human being intervantion needed

Comment

About The Project

Inspiration

At our university, attendance tracking is a real and recurring issue. Large class sizes, manual roll calls, and delayed reporting make it hard to identify students at risk early. We wanted to solve a problem we see every semester: by the time attendance issues are noticed, it is often too late for timely intervention.

This project was inspired by one core question:
How can we convert attendance from a manual checklist into an automated early-warning system?

What We Built

We built an AI-powered attendance monitoring platform that automates the full workflow:

  1. Classroom capture and session monitoring
  2. Attendance discrepancy detection
  3. Low-attendance risk identification (below 75%)
  4. Automated warning email generation
  5. Subject-level prioritization insight for each student

The platform gives faculty and admins a live dashboard view (Course -> Subject -> Class), camera/analysis workflow, historical reporting, and student-specific mail previews.

How We Built It

Our stack:

  1. Next.js + TypeScript for full-stack app architecture
  2. Prisma + PostgreSQL for structured attendance/session data
  3. Gemini API for analysis and communication intelligence
  4. Nodemailer for automated warning mail pipeline
  5. Face-detection-assisted classroom feed workflow for attendance context

Core system flow:

  1. Capture classroom feed snapshots
  2. Analyze session context and attendance signals
  3. Compare across class intervals
  4. Compute attendance percentage by student
  5. Flag students under threshold
  6. Generate personalized warning content automatically

Challenges We Faced

  1. Deployment consistency on cloud builds (Prisma client + cached dependency behavior)
  2. Runtime stability between local and production environments
  3. Handling missing/incomplete real-world data during demo scenarios
  4. Ensuring the platform remains functional even when external services are partially unavailable
  5. Designing a judge-ready UI that clearly demonstrates end-to-end automation

What We Learned

  1. Production readiness is not just coding features, it is handling edge cases and deployment behavior.
  2. AI is most useful when connected to operational workflows, not used as an isolated chatbot.
  3. Good intervention systems require explainability, not just alerts.
  4. Building for real institutional use means reliability, graceful fallbacks, and clear data visibility.

Why This Matters

This is not a hypothetical problem for us. Attendance management is one of the practical issues in our university, and this project is our attempt to make that process intelligent, proactive, and scalable.

A Simple Framing

If attendance risk is modeled as:

$$ R = \max(0, 75 - A) $$

where $A$ is a student’s attendance percentage, then our platform is designed to detect non-zero risk early and trigger intervention automatically, before academic impact compounds.

Outcome

We transformed attendance from a reactive administrative task into a proactive AI-assisted intervention pipeline.
The result is faster decisions, lower manual workload, and better support for students who need attention first.

Built With

Share this project:

Updates