Clusy

Clusy is an agent-native notebook for machine learning and data science. Instead of writing code cell by cell, you describe the outcome you want in plain English — like "fine-tune this model on my CSV" — and Clusy handles the rest. It scouts datasets, checks hardware requirements, writes the notebook, and executes it end-to-end. The example on the site shows a user asking Clusy to fine-tune DeepSeek-R1-Distill-Qwen-32B on a Spanish support sample, keeping it to one A100 hour. Clusy responds by spawning sub-agents: one to find the right dataset on HuggingFace, another to verify the model fits on the GPU, then it stitches together a full training pipeline with PEFT and bitsandbytes.

What makes Clusy stand out is how it handles complex, multi-step ML workflows. You can queue follow-up requests while it's still working — for instance, asking it to keep the adapter export small and include sanity prompts. The interface shows you its reasoning, the tools it activates (like HuggingFace search), and the plan it builds step by step. It's not just a code generator; it's an orchestrator that delegates tasks to specialized agents (data analyst, software engineer) and waits for their results before proceeding.

Clusy is built for data scientists and ML engineers who want to move faster, especially when dealing with messy CSV files or repetitive fine-tuning tasks. It's also useful for teams that need to standardize ML workflows — instead of manually setting up environments and writing boilerplate, you describe the goal and let Clusy handle the execution. If you're comfortable with the command line but tired of boilerplate, this tool could save you hours per project.