Frequently Asked Questions
What is distillation?
A process to creating high-quality data automatically that would have required thousands of human hours to manually annotate. We use this rich dataset to train task-specific SLMs that are optimised for your specific use case.
How long does it take to onboard?
The distillation process only takes a few days, and we will work with you to get onboard quickly. Timelines vary based on your company systems, data, and processes.
What type of data can be used?
We generally focus on unstructured data (text documents, i.e., PDFs, Word, raw data, etc.)
Are you secure?
Yes, we’re working with Drata to be SOC2 and ISO270001 compliant. Furthermore, because we use open-source models, we do not send your data anywhere else or use your data to train models for other customers.
Who hosts the system?
We can host on Malted AI’s secure cloud or we can deploy on your AWS or Azure.
How are you different from closed-source LLMs?
We are more customised and performant at domain-specific tasks. We provide flexible deployment and security options, and smaller models are more economic to host.
How are you different from standard developer tools?
Developer tools allow the chaining of models and components. Malted AI works with enterprises to build custom AI applications using our platform technology. We also maintain and update your solution over time.
How do you evaluate your performance?
We use domain experts or A/B tests to evaluate the systems.
Where is your technology best applied?
Our technology could be applied to multiple use cases across different industries. We specialise in regulated industries where data security and factual accuracy are a priority.
This includes finding answers and insights hidden in vast amounts of unstructured data a human would spend days looking through.
Malted AI can support operational automation, knowledge management, and the creation of specific content using traceable and consistent AI.