The announcement
Mastranet AI has joined the NVIDIA Inception Program, NVIDIA's programme for startups building products on artificial intelligence, data science and accelerated computing.
For us it means something very concrete: accelerating the development of TypeLens and Mira, and making industrial automation more accessible to the companies we work with.
What the NVIDIA Inception Program is
Inception is not an accelerator in the classic sense: it takes no equity and has no fixed-term cohort. It is a free technical and commercial support programme for companies building products on GPUs and the NVIDIA stack.
The main things it provides:
- Access to the technology: NVIDIA SDKs, libraries and frameworks, with preferential terms on hardware and software.
- Technical support: direct discussion with NVIDIA teams on architecture, training and inference optimisation.
- Training: Deep Learning Institute courses for the development team.
- Visibility and go-to-market: presence in the NVIDIA ecosystem and introductions to investors and partners.
The level of support scales with company maturity: it is designed to accompany growth rather than to reward a milestone already reached.
What it changes for TypeLens
TypeLens reads documents - delivery notes, invoices, orders, email attachments - extracts the useful fields and writes them into the ERP. Under the hood that is an inference problem: vision and language models that have to run at high volume, with response times compatible with how an office actually works.
Working on the NVIDIA stack means we can optimise the most expensive part of the product - processing time per document - and therefore keep automation viable for companies handling thousands of documents a month. If the underlying technology interests you, we cover it in what OCR is and how it works.
What it changes for Mira
Mira works on industrial troubleshooting and computer vision, and automated quality control is by definition a GPU workload. The benefit here is twofold: faster model experimentation, and lower inference cost in production - which is what determines whether a line can afford to inspect every single part or only a sample.
Why it matters if you work with us
Announcements like this risk being self-referential, so here is the part that concerns customers. Joining Inception does not change what the product does: it changes how quickly we can improve it and how cheaply we can run it. For software priced by document volume, inference cost is not a technical footnote - it is a line in the quote.
If you want to know what that means for your own documents, let's talk.