_db33a8.png)
.thumb.png.9594b08ba898a6b0d0fa1576ea2c9574.thumb.png.3097220ace298f0faaab6884cd0ea0f5.png)
This hackathon challenges participants to build functional machine learning-powered web applications — from the ground up. No shortcuts. No language model APIs. Just real models, real code, and real creativity.
Unlike typical AI hackathons, this event is about building authentic machine learning solutions — not calling an API. You'll train your own models, deploy them into actual web apps, and explain exactly how they work.
What Makes This Hackathon Different-
No LLM APIs allowed
(e.g. ChatGPT, Gemini, Claude, BERT, or any pre-trained black-box models) -
Models must be trained or fine-tuned by you
No wrappers, no hosted inference — just your own ML work -
Lightweight, interpretable, and focused apps
We value usefulness and clarity over complexity -
Judging based on transparency, originality, and execution
The more open and reproducible your project, the better
Requirements
-
To be eligible for judging and prizes, each team must submit the following via Devpost:
-
A functional web application that:
-
Uses a machine learning model trained or fine-tuned by the team
-
Includes both frontend and backend components
-
Does not use any LLM APIs, hosted language models, or pre-trained wrappers (e.g. ChatGPT, Gemini, BERT, etc.)
-
-
A written project description that includes:
-
What your app does
-
How the ML model was developed and trained
-
How it integrates into the app
-
The problem it solves and your approach
-
-
A demo video (2–5 minutes) that:
-
Shows the working app
-
Explains how the model works
-
Walks through key features and technologies used
-
-
A public code repository (GitHub or similar) that contains:
-
The full application code (frontend + backend)
-
The training code for the ML model
-
References or links to the datasets used
-
A README with setup and deployment instructions
-
-
Open Source Requirement (Mandatory):
-
All code and training workflows must be fully open-sourced
-
Submissions without open access to training code, dataset references, and app code will not be considered for prizes, regardless of quality
-
We are promoting reproducibility and transparency through open source. If your repo is private or incomplete, your submission will be disqualified
-
-
Deployment:
-
Live deployment link (e.g. Render, Vercel, Streamlit, Hugging Face Spaces)
-
-
Prizes
Best Real ML Web App (1st Place)
Best Real ML Web App (2nd Place)
Best Real ML Web App (3rd Place)
Best Documentation Award
Most Creative Idea
Devpost Achievements
Submitting to this hackathon could earn you:
Judges
Alex Huang
Product Manager, ML Tools
Dr. Rafael Torres
Professor of Computer Science
Fatima Al-Hassan
Applied Data Scientist
Judging Criteria
-
Originality
Is the idea unique, creative, or presented in a novel way? -
Technical Execution
Does the project demonstrate strong implementation of both ML and web components? -
ML Authenticity
Was the model trained or fine-tuned by the team, and does it reflect actual ML work (not just using pre-built wrappers)? -
Functionality
Does the application work end-to-end? Is it usable and reasonably bug-free? -
Usefulness
Does the project solve a real problem, serve a clear purpose, or have practical potential? -
Presentation
Is the demo video clear? Is the submission well-documented and easy to understand? -
Documentation
Is the project clearly explained? Does it include code comments, model design choices, training details, dataset source, and setup instructions? This is one of the most heavily weighted categories.
Recommended Comments
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.