top of page

AIxBio Global Sprint Hackathon | Nov 7-16, 2025

8 GPUs

25+ AI Models

3 Challenges

22 Judges

Top-tier Prizes

Models and Resources

Each team will receive:

  1. Unrestricted access to 8 H100s for 10 days

  2. $100 credits to the Nebius AI studio

  3. Training and tuning access to all models in the NVIDIA BioNeMo framework

  4. Inference-only access to select additional models from NVIDIA NIMs

​

 

A subset of available models is listed below. Visit the Nebius AI studio for a complete list.

8 H100 GPUs

allocated per team

$100 Nebius AI Studio Credits

NVIDIA BioNeMo Framework

CHALLENGES

Agentic Automation and Workflow Optimization

AI systems that act in collaboration with scientists, engineers, or clinicians to accelerate existing workflows across research, development, and operations. 

Modeling Biological and Biomedical Systems

Methods to improve representation and inference in intricate biological systems or clinical challenges, with multi-modal or biologically-informed architectures, or novel applications of foundation models.   

Molecular Optimization and Genomic Modeling 

Methods for modeling genomic sequences and molecular structures for drug development and target or biomarker identification, working with data at the genomic, transcriptomic, or protein level.

REGIONAL HACKATHONS

Not an incorporated start-up ready to commit to a 10 day sprint? Regional hackathons provide short-term access to the Nebius cloud computing platform and NVIDIA modeling resources. A great fit for AI/ML engineers looking to experiment with projects in healthcare and biology, startups looking for new team members, or anyone interested in building with the models available through NVIDIA BioNeMo and the Nebius AI Studio. ​​​​​

​

Prior experience building and training AI models is recommended. Applicants are not required to apply with a pre-existing team or project. 

Boston, MA

​Join us at LabCentral 238 on October 24th!​Boston AI researchers, engineers, and scientists will come together for a day of rapid innovation with cutting-edge computational tools. Apply by Oct 20th. 

San Francisco, CA

Join us in San Francisco on November 13th to hack on Agentic AI for biotech and clinical development, alongside our SF-based global sprint participant teams. After a day of innovation, pitch your project and celebrate the SF AIxBio community! Apply by Nov 7th. 

JUDGES

Screenshot 2025-10-10 at 11_edited.jpg

Available Models

​

All NVIDIA BioNeMo framework models are available for fine-tuning. Models from the Nebius AI Studio and NVIDIA NIMs are available inference-only.

 

A subset of available models is listed below. Visit the Nebius AI studio for a complete list of available LLMs. 

gpt-oss

openAI

OpenAI’s open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases. Available in 20b and 120b variants. 

kimi-k2-instruct

moonshotAI

A state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters

ESM-2

Meta AI

ESM-2 is a state-of-the-art protein model trained on a masked language modelling objective

arctic-embed-1

Snowflake

An embedding model that focuses on creating high-quality retrieval models optimized for performance.

Hermes-4

NousResearch

OpenAI’s open-weight models designed for powerful reasoning, agentic tasks, and versatile developer use cases. Available in 20b and 120b variants. 

GLM-4.5

moonshotAI

A state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters

Qwen3

Qwen

Qwen3 is the latest generation of LLMs in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Multiple sizes/versions available. 

Llama-3.1, 3.3

Meta

A protein language model  designed for protein sequence understanding and prediction tasks.  120M and 350M parameter sizes available.

Boltz- 2

MIT

Screenshot 2025-10-09 at 11.33.18 AM.png

Predict complex protein structures. 

Evo2

arc

Evo 2 is a biological foundation model that is able to integrate information over long genomic sequences while retaining sensitivity to single-nucleotide changes

Geneformer

Broad Institute

Geneformer generates a dense representation of a sc-RNA cell by learning co-expression patterns within single cells.

AMPLIFY

OpenAI

A protein language model  designed for protein sequence understanding and prediction tasks.  120M and 350M parameter sizes available.

Evolved 2025 is enabled by

NVIDIA_Logo_1_edited.png
Screenshot 2025-09-03 at 3.31_edited.png
bottom of page