ML Intern

Short Answer

ML Intern is an open-source AI research agent from Hugging Face that automates parts of the machine learning research loop. Built on the company’s smolagents framework, it can browse arXiv and Hugging Face Papers, find and prepare datasets from the Hub, run training scripts, and iterate on evaluations, mirroring how an ML researcher actually works.

It’s a technical tool for ML researchers and AI developers, not a general-purpose assistant.

What ML Intern Is Best For

• ML research workflows: automate the paper → dataset → train → evaluate loop
• Post-training experiments: iterate on model improvements
• Dataset workflows: discover, inspect, and reformat datasets
• Research agents: a continuous loop that behaves like a junior researcher

Practical Use Cases

• Explore and read research papers
• Find and prepare datasets
• Run training experiments
• Debug ML workflows
• Run evaluations and iterate on model experiments

Honest Limitations

• Not beginner-friendly: requires real ML knowledge
• Compute and setup can be complex: experiments need GPUs and configuration
• Results need expert review: automation speeds the loop, it doesn’t replace judgment

Alternatives Worth Knowing

• Hugging Face, the platform ML Intern builds on for models and datasets
• GLM 5.2, open-weight model you can experiment with
• Ollama, run open models locally for testing
• LM Studio, GUI for running and evaluating local models