genbio.ai
Hosted API and ModelGenerator framework to run and fine-tune the AIDO family across RNA, DNA, protein, and single-cell.
Overview
genbio.ai is the lab behind the AIDO family of multimodal foundation models, offering hosted inference and fine-tuning across several biological modalities from a single source. Because one lab develops AIDO for RNA, DNA, protein, and single-cell data, you get consistent tooling and a shared framework instead of a different stack per modality. This is the place to run or fine-tune AIDO models when you want API-based inference or a training framework rather than assembling per-modality pipelines yourself.
What you can run on genbio.ai
genbio.ai serves the AIDO model family across four modalities. AIDO.RNA is an RNA foundation model for RNA sequence and structure tasks; AIDO.DNA is a genomic language model for DNA sequence modeling; AIDO.Protein is a protein language model for sequence-to-property and representation tasks; and AIDO.Cell is a single-cell model for transcriptomic embeddings and cell-level analysis. All four are available for both hosted inference and fine-tuning. The model weights are published openly on Hugging Face; genbio.ai provides the API and training framework that run them, so you can work with the models through managed endpoints or adapt them to your own data.
Running and fine-tuning models on genbio.ai
There are two access modes. A hosted API delivers inference against AIDO.RNA, AIDO.DNA, AIDO.Protein, and AIDO.Cell without local setup — a fit for teams that want embeddings or predictions on demand. For customization, the open ModelGenerator framework supports fine-tuning and running the models on your own data and infrastructure, aimed at ML practitioners who need task-specific adaptation across RNA, DNA, protein, or single-cell problems. The shared framework across the AIDO family means a workflow built for one modality transfers cleanly to the others, which helps groups working across several biological data types at once.
Run inference on genbio.ai (4)
RNA foundation model with 1.6 billion parameters, pretrained on 42 million non-coding RNA sequences for structure prediction and RNA sequence design.
Mixture-of-experts protein language model scaling to 16 billion parameters, applied to variant effect prediction and de novo protein design.
Single-cell RNA-seq foundation model pretrained on 50 million human cells, encoding the full transcriptome for annotation and perturbation modeling.
DNA foundation model scaling an encoder-only transformer to 7 billion parameters for variant effect prediction, gene expression, and sequence design.
Fine-tune on genbio.ai (4)
RNA foundation model with 1.6 billion parameters, pretrained on 42 million non-coding RNA sequences for structure prediction and RNA sequence design.
Mixture-of-experts protein language model scaling to 16 billion parameters, applied to variant effect prediction and de novo protein design.
Single-cell RNA-seq foundation model pretrained on 50 million human cells, encoding the full transcriptome for annotation and perturbation modeling.
DNA foundation model scaling an encoder-only transformer to 7 billion parameters for variant effect prediction, gene expression, and sequence design.