Normally, to re-implement one project in multiple frameworks, developers must create spin-off codebases – that often deviate from the original and may have performance issues as a result – then test the code, discuss errors, and iterate to address them, all of which takes huge amounts of time. With Ivy, code is converted directly to the new framework, with a computation graph guaranteed to be identical to the original, massively reducing time costs. In addition, code written with Ivy will support future ML frameworks with zero changes.

Ivy’s roadmap shows that the company will add to its product in the push to unify all ML frameworks, with upcoming projects including framework-specific front-ends and verified code conversions. The company has raised initial funding and support, with immediate plans to hire talent for its expanding team.