Evolva is a genetic-validated drug discovery platform powered by a Small Language Model (SLM) trained on curated molecular datasets. It predicts pIC50 values, validates drug targets, synthesizes reaction routes, and identifies patent workarounds.

Model Description

The Evolva Molecular SLM is a domain-specific language model fine-tuned on:

ChEMBL bioactivity data (pIC50 values)
SMILES molecular representations
Target-ligand interaction data
Reaction pathway databases

Capabilities

Task	Description
pIC50 Prediction	Predict binding affinity from SMILES input
Target Validation	Validate genetic drug targets
Route Synthesis	Generate synthetic reaction pathways
Patent Workaround	Identify non-infringing molecular variants

Usage

python

from huggingface_hub import hf_hub_download

# Model will be available after Sprint 46 training completes
model_path = hf_hub_download(
    repo_id="Pippinlitli/evolva-molecular-slm",
    filename="model.gguf"
)

Inference API

Live predictions available at: https://huggingface.co/spaces/Pippinlitli/evolva-inference

Demo UI

Interactive Gradio demo: https://huggingface.co/spaces/Pippinlitli/evolva-frontend

Status

Note: The model.gguf file will be uploaded after Sprint 46 training completes. The inference API currently returns a graceful error message until the model is available.

Citation

bibtex

@misc{evolva2026,
  title={Evolva: Genetic-Validated Drug Discovery via Molecular SLM},
  author={Gudmundur Eyberg Kristjansson},
  year={2026},
  url={https://huggingface.co/Pippinlitli/evolva-molecular-slm}
}

License

Apache 2.0