What is ekam_?

EKAM is a platform for running competitions and events such as hackathons, quizzes, coding contests, and knockout tournaments, using AI to set up the event and software to actually run it. An organizer describes the event in plain English, and the system turns that description into a structured plan the README calls a Universal Blueprint, asking follow up questions for anything missing before the organizer approves it and the event goes live. The project's central idea is that most event types can be broken down into the same small set of building blocks: registration, team or individual sign up, rounds of submission and evaluation, and a final result. Because of this, the same engine can run a hackathon, an ideathon, a case competition, a coding contest, a quiz, a knockout bracket, or a paper review symposium, and it can attempt to fit unusual formats like a debate or scholarship process onto those same building blocks too. Several pieces of automation sit behind the AI layer. Constraint based solving forms balanced teams from individual sign ups based on skills and diversity, and assigns judges to teams or rounds. Machine learning is used to flag scoring anomalies, such as unusual timing or bias, and to check submitted code, PDFs, or repositories for similarity to catch possible plagiarism. A public registration page adapts its fields to the event type, includes bot protection, and enforces registration windows and capacity limits. The system deliberately never lets the AI make an irreversible decision alone. Every step that cannot be undone, such as deploying the event, forming teams, assigning judges, or announcing winners, requires an organizer to click approve first. Real time updates are pushed to organizer, participant, and judge dashboards over a persistent connection so results and bracket changes appear instantly. Under the hood, the backend is built with FastAPI, async SQLAlchemy, and PostgreSQL, the frontend uses Next.js and React, and the AI layer relies primarily on Gemini with an automatic fallback to Groq if the primary model is unavailable.