# AutoGrade - AI Paper Evaluation

> A fire-and-forget edge chain - extract → grade → qc-check - sidesteps the 150s serverless timeout by treating Postgres status as the orchestrator. Vision-LLM reads handwritten + Gujarati scripts; self-consistency voting at temp=0 makes grading deterministic; an arithmetic QC gate catches AI hallucinations before they reach the teacher.

**Category:** Vision-LLM pipeline · determinism + QC  
**Role:** Solo full-stack + AI engineer  
**Status:** Shipped  
**Year:** 2025  
**Canonical URL:** https://koratmeet.in/projects/autograde

## Summary

Web app where teachers upload a question paper once, then each student's handwritten or scanned PDF. The system returns per-question marks, deductions with evidence, a split-view PDF viewer with bounding-box highlights, and a QC verdict.

Handles real-world mess: handwriting, multi-page PDFs, mixed English/Gujarati scripts, merged sub-questions like `3(a,b,c)`, OMR-bubble roll number extraction - with an audit trail a teacher can defend to a parent.

## Problem

A teacher with 60 students × 20 questions burns 6–10 hours per exam, with inconsistent strictness across the batch and zero traceability of why marks were deducted. No off-the-shelf product handles handwritten, multi-page, mixed-script papers with auditable evidence.

## Solution

React/Vite SPA → Supabase (Postgres + RLS + Auth + Storage + Edge Functions) → Lovable AI Gateway (Gemini Pro, vision). Fire-and-forget edge chain with DB-status orchestration. react-pdf canvas rendering enables bounding-box evidence overlays.

## Architecture

QP Upload → Parse → Student PDF → Extract → Grade (parallel) → QC Check → Results

## Stack

- React 18
- Vite
- TypeScript
- Tailwind
- shadcn/ui
- Supabase
- Edge Functions (Deno)
- Lovable AI Gateway
- Gemini 2.5 / 3.1 Pro Preview (vision)
- react-pdf

## Metrics

- **Grading time:** ~90% saved
- **Per paper:** ~2 min
- **Re-runs:** Deterministic
- **Edge functions:** 4

## Key engineering decisions

### DB-as-orchestrator instead of Inngest/SQS

Work is self-bounded (one student paper) with no fan-out. Postgres status is the source of truth - inspectable with `SELECT status FROM evaluations`. The right level of complexity for the problem.

### Fire-and-forget edge chain

AI pipeline is 60–120s; edge timeout is 150s. Chaining functions via `fetch keepalive:true` decouples the long job from the HTTP request lifecycle.

### Determinism contract across every AI call

temp=0, top_p=0.1, self-consistency voting. Same paper graded twice produces the same marks - non-negotiable for teacher trust.

### Canvas (react-pdf), not iframe

Bypasses browser PDF sandboxing AND enables bounding-box overlays for evidence - required for the product UX.

## Build notes

- 4 Deno edge functions: parse-question-paper, extract, grade, qc-check - JWT-verified at the boundary, service-role for internal chaining
- Per-question parallel grading at `temp=0`, `top_p=0.1` with self-consistency voting for reproducible marks
- qc-check re-derives totals, verifies arithmetic, evaluates evidence quality, flags low-confidence questions; emits PASS / FLAG / FAIL
- react-pdf canvas (not iframe) renders pages so deductions can highlight the exact bounding box on click
- UI polls `evaluations.status` every 3s; progress bar maps pending→processing→grading→qc_check→completed
- Multi-tenant security: RLS on every table, private buckets with short-lived signed URLs, HIBP leaked-password protection

## Results

- ~90% reduction in grading time per batch
- Deterministic re-runs: identical paper → identical marks
- Zero arithmetic errors reach the teacher (QC gate)
- Every deduction defensible: click → bounding box on the rendered PDF

## What this demonstrates

- Production vision-LLM application on real handwritten scripts
- LLM determinism as a product feature, not a prompt detail
- Pragmatic distributed-systems thinking around serverless timeouts
- Evidence-linked AI UX with bounding-box overlays
- Vision LLM in production on a messy real-world dataset
- Self-consistency + arithmetic QC as guardrails - not just prompting
- Timeout-driven architecture: design the workflow around the runtime constraint
- Evidence UX: every AI decision is traceable to a region on the page

---

Built by Meet Korat, AI Automation Engineer. Contact: meetkorat903@gmail.com · https://koratmeet.in