20 years into building AI systems across Big Tech, Amund Tveit kept seeing the same issue: breakthrough research sits idle for months while teams manually port models. Atsentia turns AI papers into production-ready code across JAX, PyTorch, CUDA, ROCm, and Metal.

Most recently Principal PM for Microsoft’s Copilot Search, Tveit previously led engineering at YouTube Zürich and built large-scale production monitoring infra at Google. A repeat founder with a PhD from NTNU, 800+ citations, and five patents, he brings decades of translating research into production.

Atsentia’s moat is a proprietary translation engine that learns from the latest papers - a living library of conversion recipes that improves with each release. To protect this layer, Tveit’s recent patent filings add automated safety and compliance. He’s advancing the work publicly through active contributions to JAX and PyTorch.

The vision: make every AI breakthrough deployable instantly. Atsentia enables enterprise teams to ship cross-framework, cross-hardware models without rebuilding from scratch. If successful, it could collapse months of engineering into minutes and accelerate the AI innovation cycle.