Agentomics-ML¶

Autonomous AI agent for supervised machine learning model development on omics data

What is Agentomics-ML?¶

Agentomics-ML is an autonomous AI agent that develops machine learning models for omics data. Given a raw dataset, it produces:

A trained model ready to run inference on new data
A detailed report summarizing the model development process and evaluation metrics

Agentomics-ML works like an ML engineer:

Explores data before designing a model
Considers domain information from dataset descriptions
Chooses proper data representation (encoding, normalization, feature selection)
Designs and trains models, including custom neural networks
Works iteratively, reacting to issues like overfitting and underfitting based on validation metrics
Produces working scripts with their conda environments

Feature	Description
Any LLM	Works with OpenAI, OpenRouter, or local models via Ollama
Any Dataset	Supports classification or regression datasets in CSV format
Secure Execution	Docker containers with read-only access to code and isolated execution
Reproducible	Outputs include trained models, scripts, and conda environments

Choose the setup that works best for you:

Mode	Description	Best For
Docker + Pull Images	Fastest setup - pulls pre-built images	Getting started quickly
Docker + Local Build	Build images locally	Custom modifications
Local Mode	No Docker, uses conda directly	Development, Google Colab
Local LLMs	Run with Ollama	Privacy, offline use

# Set your API key
export OPENROUTER_API_KEY="your-key-here"

# Run the agent (pulls Docker images automatically)
./run.sh --pull-images

The agent will guide you through selecting a model, dataset, and run parameters interactively.

MIT. See the LICENSE.