We're building an integrated ecosystem for AI-powered biology
Biohub's platform accelerates the development and application of state-of-the-art AIxBio models. We help developers build biologically impactful models faster, and empower biologists to use them with ease — creating a cycle of collaboration that drives AI-powered biological discovery forward.
Feedback loop diagram
Diagram showing a feedback loop between data, models, benchmarks, and workflows connecting biologists and ML developers.
For developers
A developer toolkit to build faster for scientific impact
- Query and download multimodal, ML-ready biological datasets
- Integrate biologically relevant benchmarks to evaluate your model and compare it against other leading models
- Streamline model development and contribution with one powerful CLI
For biologists
An AI Workspace to supercharge your research pipelines
- Run models on your own data with no-code tools and guided tutorials
- Evaluate performance on biologically relevant, reproducible benchmarks
- Discover a curated set of AI models and data most relevant to your research
[ For developers ]
A developer toolkit to build faster for scientific impact
Access multimodal, ML-ready data to fuel your model development
- Access thousands of high-quality datasets from CELLxGENE, CryoET data portal, and more in one unified place
- All data is standardized to a cross-modality metadata schema for consistent querying
- Efficient access options through API and CLI
Understand where you can make the biggest impact with biologist-defined benchmarks
- Use biologist-defined tasks, metrics, and datasets to evaluate model performance
- Identify gaps where your model can make a difference
- Evaluate performance as you build with pre-built benchmarking packages
Build and share your model faster with a single, powerful CLI
- Speed up cycles of model development and evaluation with programmatic access to data, model, and benchmarking commands
- Contribute to the scientific community and submit your model for others to explore.
[ For biologists ]
An AI Workspace to supercharge your research pipelines
Compare models using biologically relevant benchmarks to confidently choose the best one for your research needs
- Our benchmarks are standardized, reproducible, and community-driven, prioritizing performance on biologically relevant tasks over standard model performance metrics.
- We’re starting with single-cell analysis tasks and expanding to more domains.
Run models on your data with flexible no-code or low-code options. Inference included.
- Each model comes with a vetted quickstart notebook
- Try the beta version of our interactive, no-code tools for single cell embedding and analysis
- Upload your own data to run with models to generate embeddings, and use interactive visualizations to analyze the results.
Explore newly-added models
Our approach
At Biohub, we’re building the technology to help scientists around the world use AI-powered biology to dramatically improve the ability to understand and manage disease.
Our strategy is to create a flywheel for scientific discovery. We do this by building new technologies, generating biological datasets, developing new AI models, and conducting frontier research that work together to engineer novel biological systems. This virtuous cycle opens entirely new and more effective pathways for understanding and advancing human health.