How do astronomers find planets light-years away? They don't see them directly. They watch starlight — and look for the faint, telltale dips that betray a planet passing in front of its star.
Now, it's your turn.
We invite high school students across India to step into the role of an astrophysicist and data scientist. Using a curated dataset sourced directly from the NASA Exoplanet Archive, your mission is to build a machine learning classification model that separates real exoplanet candidates from noise and false signals — the exact kind of problem professionals at NASA and research institutions tackle every day.
You'll receive a starter CSV file with real observational data. From there, it's on you:
🔍 Explore — understand distributions, correlations, and anomalies 🧹 Clean — handle missing values and class imbalance thoughtfully 🤖 Model — apply classification algorithms of your choice (Random Forest, XGBoost, Neural Networks — your call) 📊 Interpret — explain what your model learned and why it matters
No astrophysics background required. Just curiosity, code, and commitment :)
Most student competitions give you toy datasets and synthetic problems. This one gives you real data from a real space telescope. We designed this challenge to bridge the gap between classroom learning and genuine scientific work — and to give students a project that belongs on a college application, a research portfolio, or a GitHub that actually stands out.
🤖 Sponsor Prize — 1st Place: $300 in Featherless.ai credits Sponsored by Featherless.ai — serverless AI inference platform
🥇 Top Solutions Published — Winning models and write-ups featured on our platform's newsletter and shared with the STEM community
📜 Certificates of Excellence — Awarded to all top-scoring participants
🎓 Academic Credibility — A real, portfolio-ready project backed by NASA data — the kind that stands out to colleges and research programs
Additional awards include internship offers at Celesta, Letters of Recommendation (LOR), and social media features. All participants are eligible for a Free Featherless Premium Model. THIS IS NOT A CASH PRIZE. THE COST OF THE SUBCRIBTIONS IS GIVEN.
📍 Open to high school students (Grades 9–12) across India 👤 Solo participation or teams of up to 3 members 🆓 Free to enter
To submit to the India High School Exoplanet Data Challenge, each team must provide all of the following:
1. 💻 Code / Notebook - A clean, well-commented Jupyter Notebook (.ipynb) or Python script (.py) - Must be reproducible — someone else should be able to run it top to bottom without errors - Upload to GitHub and share the public repository link on your Devpost submission
2. 📊 Model Results - A summary table of your model's performance metrics: Accuracy, Precision, Recall, F1-Score - Confusion matrix (image or inline in notebook) - Any visualisations supporting your findings (feature importance plots, ROC curves, etc.)
3. 📝 Written Summary / Report A written explanation (500–1000 words) covering: — Your approach to EDA and data cleaning — Why you chose your model(s) — Key findings and what the model learned — How you would explain your predictions to a non-technical audience
4. 🗂️ Devpost Project Page - Fill in your project title, description, and team members on Devpost - Embed or link your GitHub repository - Upload at least one screenshot or visualisation from your notebook