For developers · Daidon · Early access
Your data. Your model. Better economics.
Daidon is for teams running a repeated AI task in production, such as ticket classification or structured extraction, on a large general-purpose model. Bring examples of the task, fine-tune a smaller open-weight model and see whether it is worth deploying before you commit.
In development · early access for AI and ML engineering teams
How Daidon works
01
Prepare
Upload CSV or JSONL examples. Daidon checks quality, duplicates and sensitive data, and splits training, validation and held-out test sets.
02
Benchmark
Measure your current model on the held-out examples against acceptance criteria you set.
03
Train
Run a managed supervised fine-tune (LoRA where supported) on an open-weight model, starting with selected Qwen models.
04
Evaluate
Compare the candidate with your baseline on task accuracy, structured-output validity, latency and estimated cost.
05
Deploy or export
Deploy the candidate, download the supported model artefacts, or reject the experiment if it does not clear the bar.
Use cases
Support ticket classification
Route and label tickets with a small model trained on your own history.
Document extraction
Turn documents into structured fields with a model tuned to your formats.
Intent detection and routing
Classify leads, requests or messages at high volume and low latency.
Good to know
- Fine-tuning is not always the answer. Knowledge that changes often is usually better served by retrieval (RAG), better prompts or ordinary code, and Daidon will say when training is not worth it.
- Every training record keeps its provenance. Unverified model outputs never become trusted training labels automatically.
- Any projected savings are estimates until a reproducible experiment measures them.
Also by New Odyssey
More for developers
Next step
Get early access to Daidon
In development · early access for AI and ML engineering teams