Hybrid Assignment Evaluation in LMS: Automation and Expert Review

With the growing number of students, grading homework becomes a bottleneck in any LMS. An instructor physically cannot review 500 submissions in 24 hours, and quality suffers. We solve this with a hybrid approach: automatic grading for quizzes and code, and an optimized interface for manual grading of essays and complex assignments. Our experience shows that up to 70% of typical assignments can be auto-graded, reducing instructor time by 3–5x. For example, in a university project, we implemented automatic code grading in Python — the instructor now grades 200 submissions in 1 hour instead of 5. Budget savings on grading reach 40% compared to manual labor, translating to an average of $8,000 saved per semester for mid-sized institutions. Development cost starts at $5,000, with typical ROI achieved in 2–4 months. We ensure scalability up to 1000 concurrent checks and provide feedback on every submission. Average development time is 2–3 weeks. In this article, we'll share how we design a grading system, what components are included, and how you can order such a turnkey development.

What Problems Does Hybrid Grading Solve?

Instructor overload. Manual grading of 50+ submissions per day leads to burnout and errors. Automation removes the routine: the instructor only reviews complex assignments, while quizzes and code are graded automatically.

Diverse assignments. A single course may include quizzes, essays, code, and files. A universal system must support all formats. We integrate grading rubrics, inline PDF annotations, and a Docker sandbox for code.

Plagiarism. Text borrowing and code copying are common. Integration with Unicheck and MOSS (or a custom engine on TF-IDF) flags suspicious submissions before the instructor even sees them.

How We Implement Manual Grading

The grading interface must minimize context switching. On one screen: the student's work on the left, evaluation form on the right. Key components we implement:

• List of ungraded submissions with filters by assignment, group, date
• View student answer (text, inline file, or link)
• Grading rubric with checkboxes (if criteria are defined)
• Score field and text comment with Markdown support
• Buttons: "Accept", "Return for revision", "Next submission"
• Inline annotations on PDF (if the work is in PDF format)

For quiz-type assignments with correct answers, we implement batch grading: the instructor sees a table of all submissions with answers and can assign scores in bulk.

How Automatic Grading Works

Quiz / closed-ended tests:

async function autoGradeQuizSubmission(submissionId) { const submission = await db.submissions.findOne(submissionId, { include: ['assignment.questions'] }); let totalPoints = 0; let earnedPoints = 0; const results = []; for (const question of submission.assignment.questions) { totalPoints += question.points; const studentAnswer = submission.answers[question.id]; const isCorrect = checkAnswer(question, studentAnswer); if (isCorrect) earnedPoints += question.points; results.push({ questionId: question.id, correct: isCorrect, studentAnswer, correctAnswer: question.correctAnswer, }); } const score = Math.round((earnedPoints / totalPoints) * submission.assignment.maxScore); await db.submissions.update(submissionId, { status: 'graded', score, autoGradeResults: results, gradedAt: new Date(), }); await notifyStudent(submission.studentId, submissionId, score); } 

Code grading via tests: For programming courses, we run student code in isolated Docker containers against a test suite. This guarantees security and reproducibility.

async function runCodeTests(submissionId, code, language, testCases) { const result = await dockerRunner.run({ image: `lms-runner-${language}:latest`, // python:3.11, node:20, etc. code, tests: testCases, timeout: 10000, // 10 seconds memoryLimit: '256m', networkDisabled: true, // No network in sandbox }); return { passed: result.passedTests, total: testCases.length, output: result.stdout, errors: result.stderr, executionTime: result.durationMs, }; } 

FROM python:3.11-slim RUN pip install pytest COPY tests/ /tests/ ENTRYPOINT ["pytest", "/tests/"] 

submitted ↓ (auto-grade or manual) reviewing ↓ instructor opens work graded → notify student returned → student receives notification, can revise ↓ resubmitted resubmitted → back to reviewing