Approach
How Gardener Works
Built for Real Biological Computing
Gardener is designed around the way modern biological analysis actually happens: data is large, compute is distributed, and decisions require expert judgment.
Instead of sending raw datasets into a cloud chatbot, Gardener separates reasoning, interaction, and computation into a privacy-preserving workflow.
Our Approach
Gardener treats HPC as the primary execution environment, not an afterthought. Datasets remain on the compute system where they belong, while the agent helps launch, monitor, and organize analysis jobs.
The workflow separates reasoning, interaction, and execution. The LLM assists with planning and coordination, the GUI keeps the user in control, and the HPC performs data-intensive computation without exposing raw data to the model.
Gardener is built to work with established community pipelines such as nf-core. Instead of relying on ad hoc commands, analyses can be launched, recorded, and repeated through trusted workflow standards.
The agent helps with execution, record-keeping, and routine coordination, while biological interpretation and scientific decisions remain with the researcher.