You give it a sentence describing a text task sort these support tickets, tag this feedback, work out what the customer is asking for and it finds or builds the training data, fine-tunes, and hosts the endpoint. pay per job, no subscription.
The data step is the only part I'd claim is interesting, mostly because it's where everything went wrong. Nearly every fine-tuning tool assumes you show up with a dataset. The people who most want a custom model don't have one and have no realistic route to getting one, so that assumption quietly excludes them. So it searches dataset repos and open-data sources against an inferred task spec, ranks on label fit rather than keyword match, checks licenses before anything enters the pipeline, and generates labelled data when nothing suitable turns up. That last one is gated hard, because ungated synthetic data is worse than no data and it took a while to accept that.
An early user broke it comprehensively and the fix is most of what this version is. Nothing reaches training without passing a blocking check. If a failure was detectable before we charged you, it isn't allowed to happen after. Every error comes back with a cause and something you can click. Sounds trivial written down. Was most of the work.
What it doesn't do: anything other than text, pick between architectures for you, or improve the model after deployment. All on the list. I'd rather say it than let the landing page imply otherwise.
Building custom AI models is still complex — datasets, cleaning, training, GPUs. Most developers give up or use generic models.
Miniclay AI lets you describe what you want and handles everything: data sourcing, cleaning, and fine-tuning.
Free users get a working notebook. Paid users get 1-click cloud fine-tuning and ready-to-use models.
Miniclay AIFine-tune custom AI models from a simple prompt.
