# Manthan — AI-Powered Competition Platform

An AI-graded competition platform with Cloud Run-hosted evaluators on Gemini 3.1 Pro. Built for scale; deployed at an Indian institution.

## Key facts

- **Domain:** EDUCATION & CIVIC ENGAGEMENT
- **Status:** Built and deployed at an Indian institution
- **Use when:** Universities, civic bodies, or corporate L&D programmes want to run scaled, AI-graded competitions.
- **Reference engagement:** An Indian institution

## The problem

Scaled competitions (university festivals, civic ideathons, corporate L&D) bottleneck on human grading. Submissions sit in queues. Feedback arrives late or not at all.

## The approach

Manthan handles registration, submission, evaluation, leaderboard, and certification end to end. Evaluation runs on Gemini 3.1 Pro with structured rubrics per competition type. Human review remains in the loop where stakes warrant it.

## The architecture

Next.js 16 frontend on Firebase Hosting. Firestore for state. Cloud Run for the evaluator service. The evaluator decouples from the registration UI so it scales independently under submission spikes.

## The outcome

A scaled competition surface that returns evaluation in minutes, not weeks. Used in production at an Indian institution; portable to any similar use case.

## ATI shape

- Predictive intelligence — Rubric-based evaluation as a predictive scoring layer
- Agentic execution — Cloud Run evaluator agents handle submissions in parallel
- Secure data — Firebase Auth + Firestore rules per competition
- Actionable outcomes — Leaderboards, feedback, and certificates without manual grading

## Tech stack

- Next.js 16
- Firebase
- Cloud Run
- Gemini 3.1 Pro
- Firestore

## Impact

- Minutes — Evaluation latency from submission to feedback
- Scaled — Decoupled evaluator handles submission spikes
- In production — Live deployment at an Indian institution

## Related

- [All accelerators](https://levent.ai/accelerators/)
- [Next: Askive](https://levent.ai/accelerators/askive/)

Metrics on this page are estimated and expected improvements describing the design intent of the accelerator. Real-client delivered metrics stay in private decks; see https://levent.ai/ai-content-policy/ for the abstraction policy.

---

**Canonical URL:** https://levent.ai/accelerators/manthan/