doubleAI Opens Early Access to Maxwell Software Optimization Model
The company claims speed gains exceeding 38 times for data pipelines and 4.5 times for AI inference kernels in workloads it evaluated.

doubleAI opened early access to Maxwell on Oct. 19, introducing a model it says can improve software performance while programs run without changing their source code.
Maxwell works on machine instructions, the commands a computer executes beneath a program’s source code. doubleAI says the model mathematically verifies its adjustments while leaving the source code unchanged.
In workloads it evaluated, the company claims speed gains of more than 38 times for data pipelines and 4.5 times for AI inference kernels, computing routines used to run AI models. Those figures reflect doubleAI’s own evaluations.
The product launch follows a $200 million Series A that had been completed by Nov. 18, 2025. Lightspeed Ventures and Bessemer Venture Partners co-led the financing. doubleAI lists NVIDIA, Dell Technologies Capital, Greenfield, Pitango and BRM among its investors.
“AI has changed how software is written, but not how it runs,” said doubleAI co-founder and CEO Amnon Shashua.