Your AI atelier
Craft fine-tuned models with CLI + web app
A complete toolkit for the full lifecycle of fine-tuning LLMs. Collect examples, rate quality, train models, and evaluate results — all in one place.
Your AI atelier
A complete toolkit for the full lifecycle of fine-tuning LLMs. Collect examples, rate quality, train models, and evaluate results — all in one place.
Use whichever fits your workflow — or both.
Fast & local
Power users and solo developers. Work locally with JSONL files, iterate quickly, and keep everything in your terminal.
Team collaboration
Invite teammates to curate training data together. Non-technical collaborators can contribute without touching the terminal.
Five steps from raw data to deployed model.
Add input/output training examples
Score quality and rewrite weak examples
Split, format, and kick off fine-tuning
A/B compare against base models
Deploy your fine-tuned model
Install the CLI, then follow the workflow.
npm install -g aitelierbrew install aitelier# Initialize a new project
ait init
# Add training examples interactively
ait add
# Check dataset health
ait stats
# Split, format, and train
ait split
ait format
ait train
# Evaluate your fine-tuned model
ait evalNine commands covering the full fine-tuning lifecycle.
| Command | Description |
|---|---|
| ait init | Initialize a new project with config and data directories |
| ait add | Add training examples interactively (input/output pairs) |
| ait rate | Review and rate examples, rewrite low-quality ones |
| ait stats | View dataset health metrics and quality distribution |
| ait split | Split dataset into training and validation sets |
| ait format | Format examples into provider-specific JSONL |
| ait train | Upload data and start a fine-tuning job |
| ait status | Check the status of running training jobs |
| ait eval | Run A/B evaluations comparing fine-tuned vs base models |
Fine-tune open-source models with LoRA. OpenAI support coming soon.
Best quality. Ideal for complex tasks that need strong reasoning and instruction following.
Great balance of quality and cost. Works well for most fine-tuning use cases.
Fast and affordable. Perfect for high-volume, simpler tasks and rapid iteration.