Featured Story
Community
AutoScientist Challenge: Bring Frontier AI to the World
Today’s AI is expensive, static, and slow to change. A handful of companies build monolithic, one-size-fits-all-models optimized for the average use case. Most models work well until your use case doesn’t fit. Then you’re on your own.
We are changing that with AutoScientist. AutoScientist makes it possible to build frontier AI in days, not months. AutoScientist co-optimizes your data and model training recipes automatically, self-improving across both until quality converges on your objective.
Where Adaptive Data shapes the inputs, AutoScientist shapes the model, running the full research loop end-to-end so you walk away with models adapted to your goal. The loop runs itself: data and recipes are co-optimized in lockstep, iterating until the model converges on the behavior you described.
This summer, don’t settle for the world you have. Build for the word you want, and be part of the rising tide.We have 10 categories we invite you to build in, and walk away with your own frontier model.
We want to set new state-of-art model performance in these categories. And release the weights to Hugging Face and Kaggle so others get to use them.
$50,000 USD Prize Pool
- 10 First Prize awards - $4,000 each (one per category)
- 10 Runner Up awards- $1,000 each (one per category)
- 30 Honorary Mentions- swag & credits
How it Works
This hackathon runs in two consecutive two-week parts across 4 weeks total. You register once and you can submit to either part, or both.
Part 1: June 8 to June 22 Categories: Finance, healthcare, math and code, legal, and marketing.
Part 2: June 23 to July 6 Categories: Science, agriculture, data visualization, language, and all other domains.
Winners announced: July 10
On the single submission form and select which part you are entering. If you are submitting to a category in both parts, you may submit the form twice, once per part.
Submission Requirements
- You must submit to one of 10 categories: language (setting state-of-art for any language supported by Adaption), finance, health care, legal, science, agriculture, data visualization and chart interpretation, marketing, math and code, all other domains.
- You must release both the adapted dataset used for finetuning, as well as the weights you trained to both Hugging Face and Kaggle.
- Relative improvement. The final model released must show a measurable percentage improvement against a baseline mode on our held out in house test sets for each category.
- Bonus points for building and releasing a demo of AutoScientist as you navigate the competition. A social post on Linkedin and X sharing your project. Must tag @adaption_ai and adaption-labs.
- Win and get recognized. Top projects share a $50,000 prize pool and are featured by Adaption across research and community channels.
What We Provide
- 1000 in credits for data adaptation and free compute for AutoScientist
- Join our discord and find collaborators.
- Sessions with our research team to get advice along the way.
Who This is For
This challenge is for startups, researchers, engineers, founders, students, and builders interested in the future of adaptive intelligence.
If you believe AI systems should evolve instead of remaining static, this challenge was built for you.
Date