Case study · EdTech
Exelynt: AI that grades every student's real GitHub commits, with zero manual grading
- Client
- Exelynt
- Industry
- EdTech
- Location
- India
The client
Exelynt is a career-focused technology-education platform offering courses in web development, full-stack engineering, data science and AI/ML. Courses run online, offline, hybrid and self-paced across five languages.
The brief: let students follow week-by-week curricula, push assignment code to GitHub and receive instant AI-graded feedback; give mentors tools to author courses and track learners; and give employers a built-in job board to hire trained candidates.
- 100%
- Automated assignment grading
- 5
- Course languages
01 · The challenge
What stood in the way
Manual grading bottleneck
Reviewing every student's assignment code by hand does not scale. Slow feedback kills learning momentum.
Verifying real skill
Employers needed confidence that graduates can actually code. Upload-based submissions are easy to plagiarize and hard to verify.
Three audiences, one platform
Students, mentors and employers each needed their own workflows: enrollment and progress, course authoring and review, hiring and job posts.
Flexible course delivery
Online, offline, hybrid and self-paced formats in five languages had to fit one course hierarchy with modules, weekly plans and prerequisites.
02 · The solution
How we built it
Automated AI code review
Students link GitHub via OAuth and push to a branch named for their module. A signature-verified webhook pulls the commit and the AI engine scores code quality, functionality, best practices and performance.
Weighted auto-grading
A weighted formula produces pass/fail against each test's pass marks and generates specific failure reasons. Results update the learner's progress analytics automatically.
Plagiarism deterrence
Grading real commits from the student's own repository, instead of file uploads, makes copied work easier to detect and less rewarding.
Pluggable AI engine
A provider abstraction supports DeepSeek, OpenAI GPT-4 and Anthropic Claude, so the grading engine can switch models as the AI landscape changes.
Full course hierarchy
Courses with modules, weekly plans, resources, prerequisites and skills, plus assessments, notifications, a blog CMS, event galleries and analytics dashboards.
Employer job board
Job posts and applications connect course graduates directly to hiring companies, closing the loop from learning to employment.
03 · The results
What changed for the client
Zero manual grading
AI review of real GitHub commits removed the grading bottleneck and gives students richer, more specific feedback.
Feedback in moments
Students get scored, actionable feedback moments after pushing code, which keeps learning momentum high.
Verifiable, job-ready graduates
Commit-based grading gives employers evidence of real coding ability, backed by an integrated job board.
A catalogue that scales
New courses, mentors and corporate partners onboard through the course hierarchy and role model, without platform changes.
Technology used
- Java 17
- Spring Boot (JHipster)
- PostgreSQL
- Liquibase
- JWT / OAuth2
- DeepSeek
- OpenAI GPT-4
- Anthropic Claude
- React
- TypeScript
- Webpack
- Jest
- AWS S3
- GitHub OAuth & Webhooks
- Prometheus
- Grafana
- SonarQube
More work like this
EduFriend
An Android app, web app and admin panel that took a coaching business from 300 offline students to 10,000 digital enrollments.
Read the case studyFinAIQ
AI financial document automation for accounting firms: six ledger integrations, GST-ready validation and a multi-tenant SaaS core.
Read the case studySikka Exchange
Retail crypto trading for India with KYC, INR payment rails, per-user custody isolation and a complete compliance back office.
Read the case studyPlanning something like Exelynt?
Tell us about your project. A senior engineer will reply within one business day with questions, a suggested approach and next steps.